From 386ccdd61e8256c8b21ee27ee2fc12438fc5ca98 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Tue, 17 Oct 2023 11:30:20 +0200 Subject: Adding upstream version 1.43.0. Signed-off-by: Daniel Baumann --- collectors/apps.plugin/apps_groups.conf | 19 +- collectors/apps.plugin/apps_plugin.c | 893 +++++++-------------- .../apps.plugin/integrations/applications.md | 113 +++ collectors/apps.plugin/integrations/user_groups.md | 113 +++ collectors/apps.plugin/integrations/users.md | 113 +++ collectors/apps.plugin/metadata.yaml | 719 +++++++---------- 6 files changed, 926 insertions(+), 1044 deletions(-) create mode 100644 collectors/apps.plugin/integrations/applications.md create mode 100644 collectors/apps.plugin/integrations/user_groups.md create mode 100644 collectors/apps.plugin/integrations/users.md (limited to 'collectors/apps.plugin') diff --git a/collectors/apps.plugin/apps_groups.conf b/collectors/apps.plugin/apps_groups.conf index 659bd0f03..9e9d83436 100644 --- a/collectors/apps.plugin/apps_groups.conf +++ b/collectors/apps.plugin/apps_groups.conf @@ -83,11 +83,12 @@ xenstat.plugin: xenstat.plugin perf.plugin: perf.plugin charts.d.plugin: *charts.d.plugin* python.d.plugin: *python.d.plugin* +systemd-journal.plugin:*systemd-journal.plugin* tc-qos-helper: *tc-qos-helper.sh* fping: fping ioping: ioping go.d.plugin: *go.d.plugin* -slabinfo.plugin: slabinfo.plugin +slabinfo.plugin: *slabinfo.plugin* ebpf.plugin: *ebpf.plugin* debugfs.plugin: *debugfs.plugin* @@ -136,7 +137,7 @@ modem: ModemManager netmanager: NetworkManager nm* systemd-networkd networkctl netplan connmand wicked* avahi-autoipd networkd-dispatcher firewall: firewalld ufw nft tor: tor -bluetooth: bluetooth bluez bluedevil obexd +bluetooth: bluetooth bluetoothd bluez bluedevil obexd # ----------------------------------------------------------------------------- # high availability and balancers @@ -159,7 +160,7 @@ chat: irssi *vines* *prosody* murmurd # ----------------------------------------------------------------------------- # monitoring -logs: ulogd* syslog* rsyslog* logrotate systemd-journald rotatelogs sysklogd metalog +logs: ulogd* syslog* rsyslog* logrotate *systemd-journal* rotatelogs sysklogd metalog nms: snmpd vnstatd smokeping zabbix* munin* mon openhpid tailon nrpe monit: monit splunk: splunkd @@ -209,7 +210,7 @@ proxmox-ve: pve* spiceproxy # ----------------------------------------------------------------------------- # containers & virtual machines -containers: lxc* docker* balena* +containers: lxc* docker* balena* containerd VMs: vbox* VBox* qemu* kvm* libvirt: virtlogd virtqemud virtstoraged virtnetworkd virtlockd virtinterfaced libvirt: virtnodedevd virtproxyd virtsecretd libvirtd @@ -238,7 +239,7 @@ dhcp: *dhcp* dhclient # ----------------------------------------------------------------------------- # name servers and clients -dns: named unbound nsd pdns_server knotd gdnsd yadifad dnsmasq systemd-resolve* pihole* avahi-daemon avahi-dnsconfd +dns: named unbound nsd pdns_server knotd gdnsd yadifad dnsmasq *systemd-resolve* pihole* avahi-daemon avahi-dnsconfd dnsdist: dnsdist # ----------------------------------------------------------------------------- @@ -271,7 +272,7 @@ backup: rsync lsyncd bacula* borg rclone # ----------------------------------------------------------------------------- # cron -cron: cron* atd anacron systemd-cron* incrond +cron: cron* atd anacron *systemd-cron* incrond # ----------------------------------------------------------------------------- # UPS @@ -319,7 +320,7 @@ airflow: *airflow* # ----------------------------------------------------------------------------- # GUI -X: X Xorg xinit xdm Xwayland xsettingsd +X: X Xorg xinit xdm Xwayland xsettingsd touchegg wayland: swaylock swayidle waypipe wayvnc kde: *kdeinit* kdm sddm plasmashell startplasma-* kwin* kwallet* krunner kactivitymanager* gnome: gnome-* gdm gconf* mutter @@ -353,11 +354,11 @@ kswapd: kswapd zswap: zswap kcompactd: kcompactd -system: systemd-* udisks* udevd* *udevd ipv6_addrconf dbus-* rtkit* +system: systemd* udisks* udevd* *udevd ipv6_addrconf dbus-* rtkit* system: mdadm acpid uuidd upowerd elogind* eudev mdev lvmpolld dmeventd system: accounts-daemon rngd haveged rasdaemon irqbalance start-stop-daemon system: supervise-daemon openrc* init runit runsvdir runsv auditd lsmd -system: abrt* nscd rtkit-daemon gpg-agent usbguard* +system: abrt* nscd rtkit-daemon gpg-agent usbguard* boltd geoclue kernel: kworker kthreadd kauditd lockd khelper kdevtmpfs khungtaskd rpciod kernel: fsnotify_mark kthrotld deferwq scsi_* kdmflush oom_reaper kdevtempfs diff --git a/collectors/apps.plugin/apps_plugin.c b/collectors/apps.plugin/apps_plugin.c index d25ae3f9b..152038968 100644 --- a/collectors/apps.plugin/apps_plugin.c +++ b/collectors/apps.plugin/apps_plugin.c @@ -265,10 +265,12 @@ struct target { uint32_t idhash; char name[MAX_NAME + 1]; - + char clean_name[MAX_NAME + 1]; // sanitized name used in chart id (need to replace at least dots) uid_t uid; gid_t gid; + bool is_other; + kernel_uint_t minflt; kernel_uint_t cminflt; kernel_uint_t majflt; @@ -782,7 +784,8 @@ static struct target *get_users_target(uid_t uid) { snprintfz(w->name, MAX_NAME, "%s", pw->pw_name); } - netdata_fix_chart_name(w->name); + strncpyz(w->clean_name, w->name, MAX_NAME); + netdata_fix_chart_name(w->clean_name); w->uid = uid; @@ -830,7 +833,8 @@ struct target *get_groups_target(gid_t gid) snprintfz(w->name, MAX_NAME, "%s", gr->gr_name); } - netdata_fix_chart_name(w->name); + strncpyz(w->clean_name, w->name, MAX_NAME); + netdata_fix_chart_name(w->clean_name); w->gid = gid; @@ -899,6 +903,14 @@ static struct target *get_apps_groups_target(const char *id, struct target *targ else // copy the id strncpyz(w->name, nid, MAX_NAME); + + // dots are used to distinguish chart type and id in streaming, so we should replace them + strncpyz(w->clean_name, w->name, MAX_NAME); + netdata_fix_chart_name(w->clean_name); + for (char *d = w->clean_name; *d; d++) { + if (*d == '.') + *d = '_'; + } strncpyz(w->compare, nid, MAX_COMPARE_NAME); size_t len = strlen(w->compare); @@ -997,6 +1009,7 @@ static int read_apps_groups_conf(const char *path, const char *file) apps_groups_default_target = get_apps_groups_target("p+!o@w#e$i^r&7*5(-i)l-o_", NULL, "other"); // match nothing if(!apps_groups_default_target) fatal("Cannot create default target"); + apps_groups_default_target->is_other = true; // allow the user to override group 'other' if(apps_groups_default_target->target) @@ -1457,17 +1470,17 @@ cleanup: netdata_log_info( "FDS_LIMITS: PID %d (%s) is using " "%0.2f %% of its fds limits, " - "open fds = %llu (" - "files = %llu, " - "pipes = %llu, " - "sockets = %llu, " - "inotifies = %llu, " - "eventfds = %llu, " - "timerfds = %llu, " - "signalfds = %llu, " - "eventpolls = %llu " - "other = %llu " - "), open fds limit = %llu, " + "open fds = %"PRIu64 "(" + "files = %"PRIu64 ", " + "pipes = %"PRIu64 ", " + "sockets = %"PRIu64", " + "inotifies = %"PRIu64", " + "eventfds = %"PRIu64", " + "timerfds = %"PRIu64", " + "signalfds = %"PRIu64", " + "eventpolls = %"PRIu64" " + "other = %"PRIu64" " + "), open fds limit = %"PRIu64", " "%s, " "original line [%s]", p->pid, p->comm, p->openfds_limits_percent, all_fds, @@ -2460,7 +2473,7 @@ static inline int debug_print_process_and_parents(struct pid_stat *p, usec_t tim for(i = 0; i < indent ;i++) buffer[i] = ' '; buffer[i] = '\0'; - fprintf(stderr, " %s %s%s (%d %s %llu" + fprintf(stderr, " %s %s%s (%d %s %"PRIu64"" , buffer , prefix , p->comm @@ -3431,8 +3444,8 @@ static void calculate_netdata_statistics(void) { // ---------------------------------------------------------------------------- // update chart dimensions -static inline void send_BEGIN(const char *type, const char *id, usec_t usec) { - fprintf(stdout, "BEGIN %s.%s %llu\n", type, id, usec); +static inline void send_BEGIN(const char *type, const char *name,const char *metric, usec_t usec) { + fprintf(stdout, "BEGIN %s.%s_%s %" PRIu64 "\n", type, name, metric, usec); } static inline void send_SET(const char *name, kernel_uint_t value) { @@ -3440,7 +3453,7 @@ static inline void send_SET(const char *name, kernel_uint_t value) { } static inline void send_END(void) { - fprintf(stdout, "END\n"); + fprintf(stdout, "END\n\n"); } void send_resource_usage_to_netdata(usec_t dt) { @@ -3518,11 +3531,11 @@ void send_resource_usage_to_netdata(usec_t dt) { } fprintf(stdout, - "BEGIN netdata.apps_cpu %llu\n" - "SET user = %llu\n" - "SET system = %llu\n" + "BEGIN netdata.apps_cpu %"PRIu64"\n" + "SET user = %"PRIu64"\n" + "SET system = %"PRIu64"\n" "END\n" - "BEGIN netdata.apps_sizes %llu\n" + "BEGIN netdata.apps_sizes %"PRIu64"\n" "SET calls = %zu\n" "SET files = %zu\n" "SET filenames = %zu\n" @@ -3549,7 +3562,7 @@ void send_resource_usage_to_netdata(usec_t dt) { ); fprintf(stdout, - "BEGIN netdata.apps_fix %llu\n" + "BEGIN netdata.apps_fix %"PRIu64"\n" "SET utime = %u\n" "SET stime = %u\n" "SET gtime = %u\n" @@ -3566,7 +3579,7 @@ void send_resource_usage_to_netdata(usec_t dt) { if(include_exited_childs) fprintf(stdout, - "BEGIN netdata.apps_children_fix %llu\n" + "BEGIN netdata.apps_children_fix %"PRIu64"\n" "SET cutime = %u\n" "SET cstime = %u\n" "SET cgtime = %u\n" @@ -3736,249 +3749,104 @@ static void normalize_utilization(struct target *root) { static void send_collected_data_to_netdata(struct target *root, const char *type, usec_t dt) { struct target *w; - send_BEGIN(type, "cpu", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, (kernel_uint_t)(w->utime * utime_fix_ratio) + (kernel_uint_t)(w->stime * stime_fix_ratio) + (kernel_uint_t)(w->gtime * gtime_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cutime * cutime_fix_ratio) + (kernel_uint_t)(w->cstime * cstime_fix_ratio) + (kernel_uint_t)(w->cgtime * cgtime_fix_ratio)):0ULL)); - } - send_END(); - - send_BEGIN(type, "cpu_user", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, (kernel_uint_t)(w->utime * utime_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cutime * cutime_fix_ratio)):0ULL)); - } - send_END(); - - send_BEGIN(type, "cpu_system", dt); for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, (kernel_uint_t)(w->stime * stime_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cstime * cstime_fix_ratio)):0ULL)); - } - send_END(); + if (unlikely(!w->exposed && !w->is_other)) + continue; - if(show_guest_time) { - send_BEGIN(type, "cpu_guest", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, (kernel_uint_t)(w->gtime * gtime_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cgtime * cgtime_fix_ratio)):0ULL)); - } + send_BEGIN(type, w->clean_name, "processes", dt); + send_SET("processes", w->processes); send_END(); - } - -#ifndef __FreeBSD__ - send_BEGIN(type, "voluntary_ctxt_switches", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->status_voluntary_ctxt_switches); - } - send_END(); - - send_BEGIN(type, "involuntary_ctxt_switches", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->status_nonvoluntary_ctxt_switches); - } - send_END(); -#endif - - send_BEGIN(type, "threads", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - send_SET(w->name, w->num_threads); - } - send_END(); - send_BEGIN(type, "processes", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - send_SET(w->name, w->processes); - } - send_END(); - -#ifndef __FreeBSD__ - send_BEGIN(type, "uptime", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, (global_uptime > w->starttime)?(global_uptime - w->starttime):0); - } - send_END(); - - if (enable_detailed_uptime_charts) { - send_BEGIN(type, "uptime_min", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->uptime_min); - } + send_BEGIN(type, w->clean_name, "threads", dt); + send_SET("threads", w->num_threads); send_END(); - send_BEGIN(type, "uptime_avg", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->uptime_sum / w->processes); - } - send_END(); + if (unlikely(!w->processes && !w->is_other)) + continue; - send_BEGIN(type, "uptime_max", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->uptime_max); - } + send_BEGIN(type, w->clean_name, "cpu_utilization", dt); + send_SET("user", (kernel_uint_t)(w->utime * utime_fix_ratio) + (include_exited_childs ? ((kernel_uint_t)(w->cutime * cutime_fix_ratio)) : 0ULL)); + send_SET("system", (kernel_uint_t)(w->stime * stime_fix_ratio) + (include_exited_childs ? ((kernel_uint_t)(w->cstime * cstime_fix_ratio)) : 0ULL)); send_END(); - } -#endif - - send_BEGIN(type, "mem", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, (w->status_vmrss > w->status_vmshared)?(w->status_vmrss - w->status_vmshared):0ULL); - } - send_END(); - - send_BEGIN(type, "rss", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->status_vmrss); - } - send_END(); - - send_BEGIN(type, "vmem", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->status_vmsize); - } - send_END(); - -#ifndef __FreeBSD__ - send_BEGIN(type, "swap", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->status_vmswap); - } - send_END(); -#endif - - send_BEGIN(type, "minor_faults", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, (kernel_uint_t)(w->minflt * minflt_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cminflt * cminflt_fix_ratio)):0ULL)); - } - send_END(); - - send_BEGIN(type, "major_faults", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, (kernel_uint_t)(w->majflt * majflt_fix_ratio) + (include_exited_childs?((kernel_uint_t)(w->cmajflt * cmajflt_fix_ratio)):0ULL)); - } - send_END(); #ifndef __FreeBSD__ - send_BEGIN(type, "lreads", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->io_logical_bytes_read); - } - send_END(); - - send_BEGIN(type, "lwrites", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->io_logical_bytes_written); - } - send_END(); -#endif - - send_BEGIN(type, "preads", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->io_storage_bytes_read); - } - send_END(); - - send_BEGIN(type, "pwrites", dt); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed && w->processes)) - send_SET(w->name, w->io_storage_bytes_written); - } - send_END(); - - if(enable_file_charts) { - send_BEGIN(type, "fds_open_limit", dt); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed && w->processes)) - send_SET(w->name, w->max_open_files_percent * 100.0); + if (enable_guest_charts) { + send_BEGIN(type, w->clean_name, "cpu_guest_utilization", dt); + send_SET("guest", (kernel_uint_t)(w->gtime * gtime_fix_ratio) + (include_exited_childs ? ((kernel_uint_t)(w->cgtime * cgtime_fix_ratio)) : 0ULL)); + send_END(); } - send_END(); - send_BEGIN(type, "fds_open", dt); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed && w->processes)) - send_SET(w->name, pid_openfds_sum(w)); - } + send_BEGIN(type, w->clean_name, "cpu_context_switches", dt); + send_SET("voluntary", w->status_voluntary_ctxt_switches); + send_SET("involuntary", w->status_nonvoluntary_ctxt_switches); send_END(); - send_BEGIN(type, "fds_files", dt); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed && w->processes)) - send_SET(w->name, w->openfds.files); - } + send_BEGIN(type, w->clean_name, "mem_private_usage", dt); + send_SET("mem", (w->status_vmrss > w->status_vmshared)?(w->status_vmrss - w->status_vmshared) : 0ULL); send_END(); +#endif - send_BEGIN(type, "fds_sockets", dt); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed && w->processes)) - send_SET(w->name, w->openfds.sockets); - } + send_BEGIN(type, w->clean_name, "mem_usage", dt); + send_SET("rss", w->status_vmrss); send_END(); - send_BEGIN(type, "fds_pipes", dt); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed && w->processes)) - send_SET(w->name, w->openfds.pipes); - } + send_BEGIN(type, w->clean_name, "vmem_usage", dt); + send_SET("vmem", w->status_vmsize); send_END(); - send_BEGIN(type, "fds_inotifies", dt); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed && w->processes)) - send_SET(w->name, w->openfds.inotifies); - } + send_BEGIN(type, w->clean_name, "mem_page_faults", dt); + send_SET("minor", (kernel_uint_t)(w->minflt * minflt_fix_ratio) + (include_exited_childs ? ((kernel_uint_t)(w->cminflt * cminflt_fix_ratio)) : 0ULL)); + send_SET("major", (kernel_uint_t)(w->majflt * majflt_fix_ratio) + (include_exited_childs ? ((kernel_uint_t)(w->cmajflt * cmajflt_fix_ratio)) : 0ULL)); send_END(); - send_BEGIN(type, "fds_eventfds", dt); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed && w->processes)) - send_SET(w->name, w->openfds.eventfds); - } +#ifndef __FreeBSD__ + send_BEGIN(type, w->clean_name, "swap_usage", dt); + send_SET("swap", w->status_vmswap); send_END(); +#endif - send_BEGIN(type, "fds_timerfds", dt); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed && w->processes)) - send_SET(w->name, w->openfds.timerfds); - } +#ifndef __FreeBSD__ + send_BEGIN(type, w->clean_name, "uptime", dt); + send_SET("uptime", (global_uptime > w->starttime) ? (global_uptime - w->starttime) : 0); send_END(); - send_BEGIN(type, "fds_signalfds", dt); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed && w->processes)) - send_SET(w->name, w->openfds.signalfds); + if (enable_detailed_uptime_charts) { + send_BEGIN(type, w->clean_name, "uptime_summary", dt); + send_SET("min", w->uptime_min); + send_SET("avg", w->processes > 0 ? w->uptime_sum / w->processes : 0); + send_SET("max", w->uptime_max); + send_END(); } - send_END(); +#endif - send_BEGIN(type, "fds_eventpolls", dt); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed && w->processes)) - send_SET(w->name, w->openfds.eventpolls); - } + send_BEGIN(type, w->clean_name, "disk_physical_io", dt); + send_SET("reads", w->io_storage_bytes_read); + send_SET("writes", w->io_storage_bytes_written); send_END(); - send_BEGIN(type, "fds_other", dt); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed && w->processes)) - send_SET(w->name, w->openfds.other); - } +#ifndef __FreeBSD__ + send_BEGIN(type, w->clean_name, "disk_logical_io", dt); + send_SET("reads", w->io_logical_bytes_read); + send_SET("writes", w->io_logical_bytes_written); send_END(); +#endif + if (enable_file_charts) { + send_BEGIN(type, w->clean_name, "fds_open_limit", dt); + send_SET("limit", w->max_open_files_percent * 100.0); + send_END(); + + send_BEGIN(type, w->clean_name, "fds_open", dt); + send_SET("files", w->openfds.files); + send_SET("sockets", w->openfds.sockets); + send_SET("pipes", w->openfds.sockets); + send_SET("inotifies", w->openfds.inotifies); + send_SET("event", w->openfds.eventfds); + send_SET("timer", w->openfds.timerfds); + send_SET("signal", w->openfds.signalfds); + send_SET("eventpolls", w->openfds.eventpolls); + send_SET("other", w->openfds.other); + send_END(); + } } } @@ -3986,312 +3854,146 @@ static void send_collected_data_to_netdata(struct target *root, const char *type // ---------------------------------------------------------------------------- // generate the charts -static void send_charts_updates_to_netdata(struct target *root, const char *type, const char *title) +static void send_charts_updates_to_netdata(struct target *root, const char *type, const char *lbl_name, const char *title) { struct target *w; - int newly_added = 0; - - for(w = root ; w ; w = w->next) { - if (w->target) continue; - if(unlikely(w->processes && (debug_enabled || w->debug_enabled))) { - struct pid_on_target *pid_on_target; - - fprintf(stderr, "apps.plugin: target '%s' has aggregated %u process%s:", w->name, w->processes, (w->processes == 1)?"":"es"); - - for(pid_on_target = w->root_pid; pid_on_target; pid_on_target = pid_on_target->next) { - fprintf(stderr, " %d", pid_on_target->pid); + if (debug_enabled) { + for (w = root; w; w = w->next) { + if (unlikely(w->debug_enabled && !w->target && w->processes)) { + struct pid_on_target *pid_on_target; + fprintf(stderr, "apps.plugin: target '%s' has aggregated %u process(es):", w->name, w->processes); + for (pid_on_target = w->root_pid; pid_on_target; pid_on_target = pid_on_target->next) { + fprintf(stderr, " %d", pid_on_target->pid); + } + fputc('\n', stderr); } - - fputc('\n', stderr); - } - - if (!w->exposed && w->processes) { - newly_added++; - w->exposed = 1; - if (debug_enabled || w->debug_enabled) - debug_log_int("%s just added - regenerating charts.", w->name); } } - // nothing more to show - if(!newly_added && show_guest_time == show_guest_time_old) return; - - // we have something new to show - // update the charts - fprintf(stdout, "CHART %s.cpu '' '%s CPU Time (100%% = 1 core)' 'percentage' cpu %s.cpu stacked 20001 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu %s\n", w->name, time_factor * RATES_DETAIL / 100, w->hidden ? "hidden" : ""); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.mem '' '%s Real Memory (w/o shared)' 'MiB' mem %s.mem stacked 20003 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute %ld %ld\n", w->name, 1L, 1024L); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.rss '' '%s Resident Set Size (w/shared)' 'MiB' mem %s.rss stacked 20004 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute %ld %ld\n", w->name, 1L, 1024L); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.vmem '' '%s Virtual Memory Size' 'MiB' mem %s.vmem stacked 20005 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute %ld %ld\n", w->name, 1L, 1024L); - } - APPS_PLUGIN_FUNCTIONS(); + for (w = root; w; w = w->next) { + if (likely(w->exposed || (!w->processes && !w->is_other))) + continue; - fprintf(stdout, "CHART %s.threads '' '%s Threads' 'threads' processes %s.threads stacked 20006 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); + w->exposed = 1; - fprintf(stdout, "CHART %s.processes '' '%s Processes' 'processes' processes %s.processes stacked 20007 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); + fprintf(stdout, "CHART %s.%s_cpu_utilization '' '%s CPU utilization (100%% = 1 core)' 'percentage' cpu %s.cpu_utilization stacked 20001 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION user '' absolute 1 %llu\n", time_factor * RATES_DETAIL / 100LLU); + fprintf(stdout, "DIMENSION system '' absolute 1 %llu\n", time_factor * RATES_DETAIL / 100LLU); #ifndef __FreeBSD__ - fprintf(stdout, "CHART %s.uptime '' '%s Carried Over Uptime' 'seconds' processes %s.uptime line 20008 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - if (enable_detailed_uptime_charts) { - fprintf(stdout, "CHART %s.uptime_min '' '%s Minimum Uptime' 'seconds' processes %s.uptime_min line 20009 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.uptime_avg '' '%s Average Uptime' 'seconds' processes %s.uptime_avg line 20010 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.uptime_max '' '%s Maximum Uptime' 'seconds' processes %s.uptime_max line 20011 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - } + if (enable_guest_charts) { + fprintf(stdout, "CHART %s.%s_cpu_guest_utilization '' '%s CPU guest utlization (100%% = 1 core)' 'percentage' cpu %s.cpu_guest_utilization line 20005 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION guest '' absolute 1 %llu\n", time_factor * RATES_DETAIL / 100LLU); + } + + fprintf(stdout, "CHART %s.%s_cpu_context_switches '' '%s CPU context switches' 'switches/s' cpu %s.cpu_context_switches stacked 20010 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION voluntary '' absolute 1 %llu\n", RATES_DETAIL); + fprintf(stdout, "DIMENSION involuntary '' absolute 1 %llu\n", RATES_DETAIL); + + fprintf(stdout, "CHART %s.%s_mem_private_usage '' '%s memory usage without shared' 'MiB' mem %s.mem_private_usage area 20050 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION mem '' absolute %ld %ld\n", 1L, 1024L); #endif - fprintf(stdout, "CHART %s.cpu_user '' '%s CPU User Time (100%% = 1 core)' 'percentage' cpu %s.cpu_user stacked 20020 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, time_factor * RATES_DETAIL / 100LLU); - } - APPS_PLUGIN_FUNCTIONS(); + fprintf(stdout, "CHART %s.%s_mem_usage '' '%s memory RSS usage' 'MiB' mem %s.mem_usage area 20055 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION rss '' absolute %ld %ld\n", 1L, 1024L); - fprintf(stdout, "CHART %s.cpu_system '' '%s CPU System Time (100%% = 1 core)' 'percentage' cpu %s.cpu_system stacked 20021 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, time_factor * RATES_DETAIL / 100LLU); - } - APPS_PLUGIN_FUNCTIONS(); + fprintf(stdout, "CHART %s.%s_mem_page_faults '' '%s memory page faults' 'pgfaults/s' mem %s.mem_page_faults stacked 20060 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION major '' absolute 1 %llu\n", RATES_DETAIL); + fprintf(stdout, "DIMENSION minor '' absolute 1 %llu\n", RATES_DETAIL); - if(show_guest_time) { - fprintf(stdout, "CHART %s.cpu_guest '' '%s CPU Guest Time (100%% = 1 core)' 'percentage' cpu %s.cpu_guest stacked 20022 %d\n", type, title, type, update_every); - for (w = root; w; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, time_factor * RATES_DETAIL / 100LLU); - } - APPS_PLUGIN_FUNCTIONS(); - } + fprintf(stdout, "CHART %s.%s_vmem_usage '' '%s virtual memory size' 'MiB' mem %s.vmem_usage line 20065 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION vmem '' absolute %ld %ld\n", 1L, 1024L); #ifndef __FreeBSD__ - fprintf(stdout, "CHART %s.voluntary_ctxt_switches '' '%s Voluntary Context Switches' 'switches/s' cpu %s.voluntary_ctxt_switches stacked 20023 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.involuntary_ctxt_switches '' '%s Involuntary Context Switches' 'switches/s' cpu %s.involuntary_ctxt_switches stacked 20024 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL); - } - APPS_PLUGIN_FUNCTIONS(); + fprintf(stdout, "CHART %s.%s_swap_usage '' '%s swap usage' 'MiB' mem %s.swap_usage area 20065 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION swap '' absolute %ld %ld\n", 1L, 1024L); #endif #ifndef __FreeBSD__ - fprintf(stdout, "CHART %s.swap '' '%s Swap Memory' 'MiB' swap %s.swap stacked 20011 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute %ld %ld\n", w->name, 1L, 1024L); - } - APPS_PLUGIN_FUNCTIONS(); -#endif - - fprintf(stdout, "CHART %s.major_faults '' '%s Major Page Faults (swap read)' 'page faults/s' swap %s.major_faults stacked 20012 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.minor_faults '' '%s Minor Page Faults' 'page faults/s' mem %s.minor_faults stacked 20011 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL); - } - APPS_PLUGIN_FUNCTIONS(); - -#ifdef __FreeBSD__ - // FIXME: same metric name as in Linux but different units. - fprintf(stdout, "CHART %s.preads '' '%s Disk Reads' 'blocks/s' disk %s.preads stacked 20002 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.pwrites '' '%s Disk Writes' 'blocks/s' disk %s.pwrites stacked 20002 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, RATES_DETAIL); - } - APPS_PLUGIN_FUNCTIONS(); + fprintf(stdout, "CHART %s.%s_disk_physical_io '' '%s disk physical IO' 'KiB/s' disk %s.disk_physical_io area 20100 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION reads '' absolute 1 %llu\n", 1024LLU * RATES_DETAIL); + fprintf(stdout, "DIMENSION writes '' absolute -1 %llu\n", 1024LLU * RATES_DETAIL); + + fprintf(stdout, "CHART %s.%s_disk_logical_io '' '%s disk logical IO' 'KiB/s' disk %s.disk_logical_io area 20105 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION reads '' absolute 1 %llu\n", 1024LLU * RATES_DETAIL); + fprintf(stdout, "DIMENSION writes '' absolute -1 %llu\n", 1024LLU * RATES_DETAIL); #else - fprintf(stdout, "CHART %s.preads '' '%s Disk Reads' 'KiB/s' disk %s.preads stacked 20002 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.pwrites '' '%s Disk Writes' 'KiB/s' disk %s.pwrites stacked 20002 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.lreads '' '%s Disk Logical Reads' 'KiB/s' disk %s.lreads stacked 20042 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.lwrites '' '%s I/O Logical Writes' 'KiB/s' disk %s.lwrites stacked 20042 %d\n", type, title, type, update_every); - for (w = root; w ; w = w->next) { - if(unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 %llu\n", w->name, 1024LLU * RATES_DETAIL); - } - APPS_PLUGIN_FUNCTIONS(); + fprintf(stdout, "CHART %s.%s_disk_physical_io '' '%s disk physical IO' 'blocks/s' disk %s.disk_physical_block_io area 20100 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION reads '' absolute 1 %llu\n", RATES_DETAIL); + fprintf(stdout, "DIMENSION writes '' absolute -1 %llu\n", RATES_DETAIL); #endif - if(enable_file_charts) { - fprintf(stdout, "CHART %s.fds_open_limit '' '%s Open File Descriptors Limit' '%%' fds %s.fds_open_limit line 20050 %d\n", type, - title, type, update_every); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 100\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.fds_open '' '%s Open File Descriptors' 'fds' fds %s.fds_open stacked 20051 %d\n", type, - title, type, update_every); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.fds_files '' '%s Open Files' 'fds' fds %s.fds_files stacked 20052 %d\n", type, - title, type, update_every); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.fds_sockets '' '%s Open Sockets' 'fds' fds %s.fds_sockets stacked 20053 %d\n", - type, title, type, update_every); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.fds_pipes '' '%s Pipes' 'fds' fds %s.fds_pipes stacked 20054 %d\n", type, - title, type, update_every); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.fds_inotifies '' '%s iNotify File Descriptors' 'fds' fds %s.fds_inotifies stacked 20055 %d\n", type, - title, type, update_every); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.fds_eventfds '' '%s Event File Descriptors' 'fds' fds %s.fds_eventfds stacked 20056 %d\n", type, - title, type, update_every); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.fds_timerfds '' '%s Timer File Descriptors' 'fds' fds %s.fds_timerfds stacked 20057 %d\n", type, - title, type, update_every); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.fds_signalfds '' '%s Signal File Descriptors' 'fds' fds %s.fds_signalfds stacked 20058 %d\n", type, - title, type, update_every); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); - - fprintf(stdout, "CHART %s.fds_eventpolls '' '%s Event Poll File Descriptors' 'fds' fds %s.fds_eventpolls stacked 20059 %d\n", type, - title, type, update_every); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); + fprintf(stdout, "CHART %s.%s_processes '' '%s processes' 'processes' processes %s.processes line 20150 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION processes '' absolute 1 1\n"); + + fprintf(stdout, "CHART %s.%s_threads '' '%s threads' 'threads' processes %s.threads line 20155 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION threads '' absolute 1 1\n"); + + if (enable_file_charts) { + fprintf(stdout, "CHART %s.%s_fds_open_limit '' '%s open file descriptors limit' '%%' fds %s.fds_open_limit line 20200 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION limit '' absolute 1 100\n"); + + fprintf(stdout, "CHART %s.%s_fds_open '' '%s open files descriptors' 'fds' fds %s.fds_open stacked 20210 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION files '' absolute 1 1\n"); + fprintf(stdout, "DIMENSION sockets '' absolute 1 1\n"); + fprintf(stdout, "DIMENSION pipes '' absolute 1 1\n"); + fprintf(stdout, "DIMENSION inotifies '' absolute 1 1\n"); + fprintf(stdout, "DIMENSION event '' absolute 1 1\n"); + fprintf(stdout, "DIMENSION timer '' absolute 1 1\n"); + fprintf(stdout, "DIMENSION signal '' absolute 1 1\n"); + fprintf(stdout, "DIMENSION eventpolls '' absolute 1 1\n"); + fprintf(stdout, "DIMENSION other '' absolute 1 1\n"); + } - fprintf(stdout, "CHART %s.fds_other '' '%s Other File Descriptors' 'fds' fds %s.fds_other stacked 20060 %d\n", type, - title, type, update_every); - for (w = root; w; w = w->next) { - if (unlikely(w->exposed)) - fprintf(stdout, "DIMENSION %s '' absolute 1 1\n", w->name); - } - APPS_PLUGIN_FUNCTIONS(); +#ifndef __FreeBSD__ + fprintf(stdout, "CHART %s.%s_uptime '' '%s uptime' 'seconds' uptime %s.uptime line 20250 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION uptime '' absolute 1 1\n"); + + if (enable_detailed_uptime_charts) { + fprintf(stdout, "CHART %s.%s_uptime_summary '' '%s uptime summary' 'seconds' uptime %s.uptime_summary area 20255 %d\n", type, w->clean_name, title, type, update_every); + fprintf(stdout, "CLABEL '%s' '%s' 0\n", lbl_name, w->name); + fprintf(stdout, "CLABEL_COMMIT\n"); + fprintf(stdout, "DIMENSION min '' absolute 1 1\n"); + fprintf(stdout, "DIMENSION avg '' absolute 1 1\n"); + fprintf(stdout, "DIMENSION max '' absolute 1 1\n"); + } +#endif } } - #ifndef __FreeBSD__ static void send_proc_states_count(usec_t dt) { @@ -4310,7 +4012,7 @@ static void send_proc_states_count(usec_t dt) } // send process state count - send_BEGIN("system", "processes_state", dt); + fprintf(stdout, "BEGIN system.processes_state %" PRIu64 "\n", dt); for (proc_state i = PROC_STATUS_RUNNING; i < PROC_STATUS_END; i++) { send_SET(proc_states[i], proc_state_count[i]); } @@ -4575,7 +4277,7 @@ static int check_capabilities() { } #endif -static netdata_mutex_t mutex = NETDATA_MUTEX_INITIALIZER; +static netdata_mutex_t apps_and_stdout_mutex = NETDATA_MUTEX_INITIALIZER; #define PROCESS_FILTER_CATEGORY "category:" #define PROCESS_FILTER_USER "user:" @@ -4629,8 +4331,8 @@ static void get_MemTotal(void) { } static void apps_plugin_function_processes_help(const char *transaction) { - pluginsd_function_result_begin_to_stdout(transaction, HTTP_RESP_OK, "text/plain", now_realtime_sec() + 3600); - fprintf(stdout, "%s", + BUFFER *wb = buffer_create(0, NULL); + buffer_sprintf(wb, "%s", "apps.plugin / processes\n" "\n" "Function `processes` presents all the currently running processes of the system.\n" @@ -4660,7 +4362,9 @@ static void apps_plugin_function_processes_help(const char *transaction) { "\n" "Filters can be combined. Each filter can be given only one time.\n" ); - pluginsd_function_result_end_to_stdout(); + + pluginsd_function_result_to_stdout(transaction, HTTP_RESP_OK, "text/plain", now_realtime_sec() + 3600, wb); + buffer_free(wb); } #define add_value_field_llu_with_max(wb, key, value) do { \ @@ -4675,7 +4379,7 @@ static void apps_plugin_function_processes_help(const char *transaction) { buffer_json_add_array_item_double(wb, _tmp); \ } while(0) -static void function_processes(const char *transaction, char *function __maybe_unused, char *line_buffer __maybe_unused, int line_max __maybe_unused, int timeout __maybe_unused) { +static void function_processes(const char *transaction, char *function __maybe_unused, int timeout __maybe_unused, bool *cancelled __maybe_unused) { struct pid_stat *p; char *words[PLUGINSD_MAX_WORDS] = { NULL }; @@ -4696,21 +4400,24 @@ static void function_processes(const char *transaction, char *function __maybe_u if(!category && strncmp(keyword, PROCESS_FILTER_CATEGORY, strlen(PROCESS_FILTER_CATEGORY)) == 0) { category = find_target_by_name(apps_groups_root_target, &keyword[strlen(PROCESS_FILTER_CATEGORY)]); if(!category) { - pluginsd_function_json_error(transaction, HTTP_RESP_BAD_REQUEST, "No category with that name found."); + pluginsd_function_json_error_to_stdout(transaction, HTTP_RESP_BAD_REQUEST, + "No category with that name found."); return; } } else if(!user && strncmp(keyword, PROCESS_FILTER_USER, strlen(PROCESS_FILTER_USER)) == 0) { user = find_target_by_name(users_root_target, &keyword[strlen(PROCESS_FILTER_USER)]); if(!user) { - pluginsd_function_json_error(transaction, HTTP_RESP_BAD_REQUEST, "No user with that name found."); + pluginsd_function_json_error_to_stdout(transaction, HTTP_RESP_BAD_REQUEST, + "No user with that name found."); return; } } else if(strncmp(keyword, PROCESS_FILTER_GROUP, strlen(PROCESS_FILTER_GROUP)) == 0) { group = find_target_by_name(groups_root_target, &keyword[strlen(PROCESS_FILTER_GROUP)]); if(!group) { - pluginsd_function_json_error(transaction, HTTP_RESP_BAD_REQUEST, "No group with that name found."); + pluginsd_function_json_error_to_stdout(transaction, HTTP_RESP_BAD_REQUEST, + "No group with that name found."); return; } } @@ -4736,13 +4443,12 @@ static void function_processes(const char *transaction, char *function __maybe_u else { char msg[PLUGINSD_LINE_MAX]; snprintfz(msg, PLUGINSD_LINE_MAX, "Invalid parameter '%s'", keyword); - pluginsd_function_json_error(transaction, HTTP_RESP_BAD_REQUEST, msg); + pluginsd_function_json_error_to_stdout(transaction, HTTP_RESP_BAD_REQUEST, msg); return; } } time_t expires = now_realtime_sec() + update_every; - pluginsd_function_result_begin_to_stdout(transaction, HTTP_RESP_OK, "application/json", expires); unsigned int cpu_divisor = time_factor * RATES_DETAIL / 100; unsigned int memory_divisor = 1024; @@ -5096,13 +4802,13 @@ static void function_processes(const char *transaction, char *function __maybe_u RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 2, "KiB/s", LReads_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_SUM, RRDF_FIELD_FILTER_RANGE, - RRDF_FIELD_OPTS_VISIBLE, NULL); + RRDF_FIELD_OPTS_NONE, NULL); buffer_rrdf_table_add_field(wb, field_id++, "LWrites", "Logical I/O Writes", RRDF_FIELD_TYPE_BAR_WITH_INTEGER, RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 2, "KiB/s", LWrites_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_SUM, RRDF_FIELD_FILTER_RANGE, - RRDF_FIELD_OPTS_VISIBLE, NULL); + RRDF_FIELD_OPTS_NONE, NULL); #endif // I/O calls @@ -5110,12 +4816,12 @@ static void function_processes(const char *transaction, char *function __maybe_u RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 2, "calls/s", RCalls_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_SUM, RRDF_FIELD_FILTER_RANGE, - RRDF_FIELD_OPTS_VISIBLE, NULL); + RRDF_FIELD_OPTS_NONE, NULL); buffer_rrdf_table_add_field(wb, field_id++, "WCalls", "I/O Write Calls", RRDF_FIELD_TYPE_BAR_WITH_INTEGER, RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 2, "calls/s", WCalls_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_SUM, RRDF_FIELD_FILTER_RANGE, - RRDF_FIELD_OPTS_VISIBLE, NULL); + RRDF_FIELD_OPTS_NONE, NULL); // minor page faults buffer_rrdf_table_add_field(wb, field_id++, "MinFlt", "Minor Page Faults/s", RRDF_FIELD_TYPE_BAR_WITH_INTEGER, @@ -5153,7 +4859,7 @@ static void function_processes(const char *transaction, char *function __maybe_u RRDF_FIELD_TYPE_BAR_WITH_INTEGER, RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 2, "pgflts/s", TMajFlt_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_SUM, RRDF_FIELD_FILTER_RANGE, - RRDF_FIELD_OPTS_VISIBLE, NULL); + RRDF_FIELD_OPTS_NONE, NULL); // open file descriptors buffer_rrdf_table_add_field(wb, field_id++, "FDsLimitPercent", "Percentage of Open Descriptors vs Limits", @@ -5165,24 +4871,24 @@ static void function_processes(const char *transaction, char *function __maybe_u RRDF_FIELD_TYPE_BAR_WITH_INTEGER, RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 0, "fds", FDs_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_SUM, RRDF_FIELD_FILTER_RANGE, - RRDF_FIELD_OPTS_VISIBLE, NULL); + RRDF_FIELD_OPTS_NONE, NULL); buffer_rrdf_table_add_field(wb, field_id++, "Files", "Open Files", RRDF_FIELD_TYPE_BAR_WITH_INTEGER, RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 0, "fds", Files_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_SUM, RRDF_FIELD_FILTER_RANGE, - RRDF_FIELD_OPTS_VISIBLE, NULL); + RRDF_FIELD_OPTS_NONE, NULL); buffer_rrdf_table_add_field(wb, field_id++, "Pipes", "Open Pipes", RRDF_FIELD_TYPE_BAR_WITH_INTEGER, RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 0, "fds", Pipes_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_SUM, RRDF_FIELD_FILTER_RANGE, - RRDF_FIELD_OPTS_VISIBLE, NULL); + RRDF_FIELD_OPTS_NONE, NULL); buffer_rrdf_table_add_field(wb, field_id++, "Sockets", "Open Sockets", RRDF_FIELD_TYPE_BAR_WITH_INTEGER, RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 0, "fds", Sockets_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_SUM, RRDF_FIELD_FILTER_RANGE, - RRDF_FIELD_OPTS_VISIBLE, NULL); + RRDF_FIELD_OPTS_NONE, NULL); buffer_rrdf_table_add_field(wb, field_id++, "iNotiFDs", "Open iNotify Descriptors", RRDF_FIELD_TYPE_BAR_WITH_INTEGER, RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 0, "fds", iNotiFDs_max, RRDF_FIELD_SORT_DESCENDING, @@ -5219,12 +4925,12 @@ static void function_processes(const char *transaction, char *function __maybe_u RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 0, "processes", Processes_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_SUM, RRDF_FIELD_FILTER_RANGE, - RRDF_FIELD_OPTS_VISIBLE, NULL); + RRDF_FIELD_OPTS_NONE, NULL); buffer_rrdf_table_add_field(wb, field_id++, "Threads", "Threads", RRDF_FIELD_TYPE_BAR_WITH_INTEGER, RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_NUMBER, 0, "threads", Threads_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_SUM, RRDF_FIELD_FILTER_RANGE, - RRDF_FIELD_OPTS_VISIBLE, NULL); + RRDF_FIELD_OPTS_NONE, NULL); buffer_rrdf_table_add_field(wb, field_id++, "Uptime", "Uptime in seconds", RRDF_FIELD_TYPE_DURATION, RRDF_FIELD_VISUAL_BAR, RRDF_FIELD_TRANSFORM_DURATION_S, 2, "seconds", Uptime_max, RRDF_FIELD_SORT_DESCENDING, NULL, RRDF_FIELD_SUMMARY_MAX, @@ -5520,69 +5226,13 @@ static void function_processes(const char *transaction, char *function __maybe_u buffer_json_member_add_time_t(wb, "expires", expires); buffer_json_finalize(wb); - fwrite(buffer_tostring(wb), buffer_strlen(wb), 1, stdout); - buffer_free(wb); + pluginsd_function_result_to_stdout(transaction, HTTP_RESP_OK, "application/json", expires, wb); - pluginsd_function_result_end_to_stdout(); + buffer_free(wb); } static bool apps_plugin_exit = false; -static void *reader_main(void *arg __maybe_unused) { - char buffer[PLUGINSD_LINE_MAX + 1]; - - char *s = NULL; - while(!apps_plugin_exit && (s = fgets(buffer, PLUGINSD_LINE_MAX, stdin))) { - - char *words[PLUGINSD_MAX_WORDS] = { NULL }; - size_t num_words = quoted_strings_splitter_pluginsd(buffer, words, PLUGINSD_MAX_WORDS); - - const char *keyword = get_word(words, num_words, 0); - - if(keyword && strcmp(keyword, PLUGINSD_KEYWORD_FUNCTION) == 0) { - char *transaction = get_word(words, num_words, 1); - char *timeout_s = get_word(words, num_words, 2); - char *function = get_word(words, num_words, 3); - - if(!transaction || !*transaction || !timeout_s || !*timeout_s || !function || !*function) { - netdata_log_error("Received incomplete %s (transaction = '%s', timeout = '%s', function = '%s'). Ignoring it.", - keyword, - transaction?transaction:"(unset)", - timeout_s?timeout_s:"(unset)", - function?function:"(unset)"); - } - else { - int timeout = str2i(timeout_s); - if(timeout <= 0) timeout = PLUGINS_FUNCTIONS_TIMEOUT_DEFAULT; - -// internal_error(true, "Received function '%s', transaction '%s', timeout %d", function, transaction, timeout); - - netdata_mutex_lock(&mutex); - - if(strncmp(function, "processes", strlen("processes")) == 0) - function_processes(transaction, function, buffer, PLUGINSD_LINE_MAX + 1, timeout); - else - pluginsd_function_json_error(transaction, HTTP_RESP_NOT_FOUND, "No function with this name found in apps.plugin."); - - fflush(stdout); - netdata_mutex_unlock(&mutex); - -// internal_error(true, "Done with function '%s', transaction '%s', timeout %d", function, transaction, timeout); - } - } - else - netdata_log_error("Received unknown command: %s", keyword?keyword:"(unset)"); - } - - if(!s || feof(stdin) || ferror(stdin)) { - apps_plugin_exit = true; - netdata_log_error("Received error on stdin."); - } - - exit(1); - return NULL; -} - int main(int argc, char **argv) { // debug_flags = D_PROCFILE; stderror = stderr; @@ -5601,6 +5251,8 @@ int main(int argc, char **argv) { error_log_errors_per_period = 100; error_log_throttle_period = 3600; + log_set_global_severity_for_external_plugins(); + bool send_resource_usage = true; { const char *s = getenv("NETDATA_INTERNALS_MONITORING"); @@ -5686,10 +5338,17 @@ int main(int argc, char **argv) { all_pids = callocz(sizeof(struct pid_stat *), (size_t) pid_max + 1); - netdata_thread_t reader_thread; - netdata_thread_create(&reader_thread, "APPS_READER", NETDATA_THREAD_OPTION_DONT_LOG, reader_main, NULL); - netdata_mutex_lock(&mutex); + // ------------------------------------------------------------------------ + // the event loop for functions + + struct functions_evloop_globals *wg = + functions_evloop_init(1, "APPS", &apps_and_stdout_mutex, &apps_plugin_exit); + + functions_evloop_add_function(wg, "processes", function_processes, PLUGINS_FUNCTIONS_TIMEOUT_DEFAULT); + + // ------------------------------------------------------------------------ + netdata_mutex_lock(&apps_and_stdout_mutex); APPS_PLUGIN_GLOBAL_FUNCTIONS(); usec_t step = update_every * USEC_PER_SEC; @@ -5697,7 +5356,7 @@ int main(int argc, char **argv) { heartbeat_t hb; heartbeat_init(&hb); for(; !apps_plugin_exit ; global_iterations_counter++) { - netdata_mutex_unlock(&mutex); + netdata_mutex_unlock(&apps_and_stdout_mutex); #ifdef NETDATA_PROFILING #warning "compiling for profiling" @@ -5708,17 +5367,15 @@ int main(int argc, char **argv) { #else usec_t dt = heartbeat_next(&hb, step); #endif - netdata_mutex_lock(&mutex); + netdata_mutex_lock(&apps_and_stdout_mutex); struct pollfd pollfd = { .fd = fileno(stdout), .events = POLLERR }; if (unlikely(poll(&pollfd, 1, 0) < 0)) { - netdata_mutex_unlock(&mutex); - netdata_thread_cancel(reader_thread); + netdata_mutex_unlock(&apps_and_stdout_mutex); fatal("Cannot check if a pipe is available"); } if (unlikely(pollfd.revents & POLLERR)) { - netdata_mutex_unlock(&mutex); - netdata_thread_cancel(reader_thread); + netdata_mutex_unlock(&apps_and_stdout_mutex); fatal("Received error on read pipe."); } @@ -5728,8 +5385,7 @@ int main(int argc, char **argv) { if(!collect_data_for_all_processes()) { netdata_log_error("Cannot collect /proc data for running processes. Disabling apps.plugin..."); printf("DISABLE\n"); - netdata_mutex_unlock(&mutex); - netdata_thread_cancel(reader_thread); + netdata_mutex_unlock(&apps_and_stdout_mutex); exit(1); } @@ -5743,21 +5399,18 @@ int main(int argc, char **argv) { send_proc_states_count(dt); #endif - // this is smart enough to show only newly added apps, when needed - send_charts_updates_to_netdata(apps_groups_root_target, "apps", "Apps"); - if(likely(enable_users_charts)) - send_charts_updates_to_netdata(users_root_target, "users", "Users"); + send_charts_updates_to_netdata(apps_groups_root_target, "app", "app_group", "Apps"); + send_collected_data_to_netdata(apps_groups_root_target, "app", dt); - if(likely(enable_groups_charts)) - send_charts_updates_to_netdata(groups_root_target, "groups", "User Groups"); - - send_collected_data_to_netdata(apps_groups_root_target, "apps", dt); - - if(likely(enable_users_charts)) - send_collected_data_to_netdata(users_root_target, "users", dt); + if (enable_users_charts) { + send_charts_updates_to_netdata(users_root_target, "user", "user", "Users"); + send_collected_data_to_netdata(users_root_target, "user", dt); + } - if(likely(enable_groups_charts)) - send_collected_data_to_netdata(groups_root_target, "groups", dt); + if (enable_groups_charts) { + send_charts_updates_to_netdata(groups_root_target, "usergroup", "user_group", "User Groups"); + send_collected_data_to_netdata(groups_root_target, "usergroup", dt); + } fflush(stdout); @@ -5765,5 +5418,5 @@ int main(int argc, char **argv) { debug_log("done Loop No %zu", global_iterations_counter); } - netdata_mutex_unlock(&mutex); + netdata_mutex_unlock(&apps_and_stdout_mutex); } diff --git a/collectors/apps.plugin/integrations/applications.md b/collectors/apps.plugin/integrations/applications.md new file mode 100644 index 000000000..f4bbc8733 --- /dev/null +++ b/collectors/apps.plugin/integrations/applications.md @@ -0,0 +1,113 @@ + + +# Applications + + + + + +Plugin: apps.plugin +Module: apps + + + +## Overview + +Monitor Applications for optimal software performance and resource usage. + + + +This collector is supported on all platforms. + +This collector supports collecting metrics from multiple instances of this integration, including remote instances. + + +### Default Behavior + +#### Auto-Detection + +This integration doesn't support auto-detection. + +#### Limits + +The default configuration for this integration does not impose any limits on data collection. + +#### Performance Impact + +The default configuration for this integration is not expected to impose a significant performance impact on the system. + + +## Metrics + +Metrics grouped by *scope*. + +The scope defines the instance that the metric belongs to. An instance is uniquely identified by a set of labels. + + + +### Per applications group + +These metrics refer to the application group. + +Labels: + +| Label | Description | +|:-----------|:----------------| +| app_group | The name of the group defined in the configuration. | + +Metrics: + +| Metric | Dimensions | Unit | +|:------|:----------|:----| +| app_group.cpu_utilization | user, system | percentage | +| app_group.cpu_guest_utilization | guest | percentage | +| app_group.cpu_context_switches | voluntary, involuntary | switches/s | +| app_group.mem_usage | rss | MiB | +| app_group.mem_private_usage | mem | MiB | +| app_group.vmem_usage | vmem | MiB | +| app_group.mem_page_faults | minor, major | pgfaults/s | +| app_group.swap_usage | swap | MiB | +| app_group.disk_physical_io | reads, writes | KiB/s | +| app_group.disk_logical_io | reads, writes | KiB/s | +| app_group.processes | processes | processes | +| app_group.threads | threads | threads | +| app_group.fds_open_limit | limit | percentage | +| app_group.fds_open | files, sockets, pipes, inotifies, event, timer, signal, eventpolls, other | fds | +| app_group.uptime | uptime | seconds | +| app_group.uptime_summary | min, avg, max | seconds | + + + +## Alerts + +There are no alerts configured by default for this integration. + + +## Setup + +### Prerequisites + +No action required. + +### Configuration + +#### File + +There is no configuration file. +#### Options + + + +There are no configuration options. + +#### Examples +There are no configuration examples. + + diff --git a/collectors/apps.plugin/integrations/user_groups.md b/collectors/apps.plugin/integrations/user_groups.md new file mode 100644 index 000000000..6f56d7be6 --- /dev/null +++ b/collectors/apps.plugin/integrations/user_groups.md @@ -0,0 +1,113 @@ + + +# User Groups + + + + + +Plugin: apps.plugin +Module: groups + + + +## Overview + +This integration monitors resource utilization on a user groups context. + + + +This collector is supported on all platforms. + +This collector supports collecting metrics from multiple instances of this integration, including remote instances. + + +### Default Behavior + +#### Auto-Detection + +This integration doesn't support auto-detection. + +#### Limits + +The default configuration for this integration does not impose any limits on data collection. + +#### Performance Impact + +The default configuration for this integration is not expected to impose a significant performance impact on the system. + + +## Metrics + +Metrics grouped by *scope*. + +The scope defines the instance that the metric belongs to. An instance is uniquely identified by a set of labels. + + + +### Per user group + +These metrics refer to the user group. + +Labels: + +| Label | Description | +|:-----------|:----------------| +| user_group | The name of the user group. | + +Metrics: + +| Metric | Dimensions | Unit | +|:------|:----------|:----| +| usergroup.cpu_utilization | user, system | percentage | +| usergroup.cpu_guest_utilization | guest | percentage | +| usergroup.cpu_context_switches | voluntary, involuntary | switches/s | +| usergroup.mem_usage | rss | MiB | +| usergroup.mem_private_usage | mem | MiB | +| usergroup.vmem_usage | vmem | MiB | +| usergroup.mem_page_faults | minor, major | pgfaults/s | +| usergroup.swap_usage | swap | MiB | +| usergroup.disk_physical_io | reads, writes | KiB/s | +| usergroup.disk_logical_io | reads, writes | KiB/s | +| usergroup.processes | processes | processes | +| usergroup.threads | threads | threads | +| usergroup.fds_open_limit | limit | percentage | +| usergroup.fds_open | files, sockets, pipes, inotifies, event, timer, signal, eventpolls, other | fds | +| usergroup.uptime | uptime | seconds | +| usergroup.uptime_summary | min, avg, max | seconds | + + + +## Alerts + +There are no alerts configured by default for this integration. + + +## Setup + +### Prerequisites + +No action required. + +### Configuration + +#### File + +There is no configuration file. +#### Options + + + +There are no configuration options. + +#### Examples +There are no configuration examples. + + diff --git a/collectors/apps.plugin/integrations/users.md b/collectors/apps.plugin/integrations/users.md new file mode 100644 index 000000000..f325f05f6 --- /dev/null +++ b/collectors/apps.plugin/integrations/users.md @@ -0,0 +1,113 @@ + + +# Users + + + + + +Plugin: apps.plugin +Module: users + + + +## Overview + +This integration monitors resource utilization on a user context. + + + +This collector is supported on all platforms. + +This collector supports collecting metrics from multiple instances of this integration, including remote instances. + + +### Default Behavior + +#### Auto-Detection + +This integration doesn't support auto-detection. + +#### Limits + +The default configuration for this integration does not impose any limits on data collection. + +#### Performance Impact + +The default configuration for this integration is not expected to impose a significant performance impact on the system. + + +## Metrics + +Metrics grouped by *scope*. + +The scope defines the instance that the metric belongs to. An instance is uniquely identified by a set of labels. + + + +### Per user + +These metrics refer to the user. + +Labels: + +| Label | Description | +|:-----------|:----------------| +| user | The name of the user. | + +Metrics: + +| Metric | Dimensions | Unit | +|:------|:----------|:----| +| user.cpu_utilization | user, system | percentage | +| user.cpu_guest_utilization | guest | percentage | +| user.cpu_context_switches | voluntary, involuntary | switches/s | +| user.mem_usage | rss | MiB | +| user.mem_private_usage | mem | MiB | +| user.vmem_usage | vmem | MiB | +| user.mem_page_faults | minor, major | pgfaults/s | +| user.swap_usage | swap | MiB | +| user.disk_physical_io | reads, writes | KiB/s | +| user.disk_logical_io | reads, writes | KiB/s | +| user.processes | processes | processes | +| user.threads | threads | threads | +| user.fds_open_limit | limit | percentage | +| user.fds_open | files, sockets, pipes, inotifies, event, timer, signal, eventpolls, other | fds | +| user.uptime | uptime | seconds | +| user.uptime_summary | min, avg, max | seconds | + + + +## Alerts + +There are no alerts configured by default for this integration. + + +## Setup + +### Prerequisites + +No action required. + +### Configuration + +#### File + +There is no configuration file. +#### Options + + + +There are no configuration options. + +#### Examples +There are no configuration examples. + + diff --git a/collectors/apps.plugin/metadata.yaml b/collectors/apps.plugin/metadata.yaml index 9794a5ea2..f24160ba7 100644 --- a/collectors/apps.plugin/metadata.yaml +++ b/collectors/apps.plugin/metadata.yaml @@ -67,160 +67,123 @@ modules: description: "" availability: [] scopes: - - name: global - description: "" - labels: [] + - name: applications group + description: These metrics refer to the application group. + labels: + - name: app_group + description: The name of the group defined in the configuration. metrics: - - name: apps.cpu - description: Apps CPU Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.cpu_user - description: Apps CPU User Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.cpu_system - description: Apps CPU System Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.cpu_guest - description: Apps CPU Guest Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.mem - description: Apps Real Memory (w/o shared) - unit: "MiB" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.rss - description: Apps Resident Set Size (w/shared) - unit: "MiB" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.vmem - description: Apps Virtual Memory Size - unit: "MiB" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.swap - description: Apps Swap Memory - unit: "MiB" + - name: app_group.cpu_utilization + description: Apps CPU utilization (100% = 1 core) + unit: percentage chart_type: stacked dimensions: - - name: a dimension per app group - - name: apps.major_faults - description: Apps Major Page Faults (swap read) - unit: "page faults/s" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.minor_faults - description: Apps Minor Page Faults (swap read) - unit: "page faults/s" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.preads - description: Apps Disk Reads - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.pwrites - description: Apps Disk Writes - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.lreads - description: Apps Disk Logical Reads - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.lwrites - description: Apps I/O Logical Writes - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.threads - description: Apps Threads - unit: "threads" - chart_type: stacked + - name: user + - name: system + - name: app_group.cpu_guest_utilization + description: Apps CPU guest utilization (100% = 1 core) + unit: percentage + chart_type: line dimensions: - - name: a dimension per app group - - name: apps.processes - description: Apps Processes - unit: "processes" + - name: guest + - name: app_group.cpu_context_switches + description: Apps CPU context switches + unit: switches/s chart_type: stacked dimensions: - - name: a dimension per app group - - name: apps.voluntary_ctxt_switches - description: Apps Voluntary Context Switches - unit: "processes" - chart_type: stacked + - name: voluntary + - name: involuntary + - name: app_group.mem_usage + description: Apps memory RSS usage + unit: MiB + chart_type: line dimensions: - - name: a dimension per app group - - name: apps.involuntary_ctxt_switches - description: Apps Involuntary Context Switches - unit: "processes" + - name: rss + - name: app_group.mem_private_usage + description: Apps memory usage without shared + unit: MiB chart_type: stacked dimensions: - - name: a dimension per app group - - name: apps.uptime - description: Apps Carried Over Uptime - unit: "seconds" + - name: mem + - name: app_group.vmem_usage + description: Apps virtual memory size + unit: MiB chart_type: line dimensions: - - name: a dimension per app group - - name: apps.uptime_min - description: Apps Minimum Uptime - unit: "seconds" + - name: vmem + - name: app_group.mem_page_faults + description: Apps memory page faults + unit: pgfaults/s + chart_type: stacked + dimensions: + - name: minor + - name: major + - name: app_group.swap_usage + description: Apps swap usage + unit: MiB + chart_type: area + dimensions: + - name: swap + - name: app_group.disk_physical_io + description: Apps disk physical IO + unit: KiB/s + chart_type: area + dimensions: + - name: reads + - name: writes + - name: app_group.disk_logical_io + description: Apps disk logical IO + unit: KiB/s + chart_type: area + dimensions: + - name: reads + - name: writes + - name: app_group.processes + description: Apps processes + unit: processes chart_type: line dimensions: - - name: a dimension per app group - - name: apps.uptime_avg - description: Apps Average Uptime - unit: "seconds" + - name: processes + - name: app_group.threads + description: Apps threads + unit: threads chart_type: line dimensions: - - name: a dimension per app group - - name: apps.uptime_max - description: Apps Maximum Uptime - unit: "seconds" + - name: threads + - name: app_group.fds_open_limit + description: Apps open file descriptors limit + unit: percentage chart_type: line dimensions: - - name: a dimension per app group - - name: apps.files - description: Apps Open Files - unit: "open files" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: apps.sockets - description: Apps Open Sockets - unit: "open sockets" - chart_type: stacked + - name: limit + - name: app_group.fds_open + description: Apps open file descriptors + unit: fds + chart_type: stacked + dimensions: + - name: files + - name: sockets + - name: pipes + - name: inotifies + - name: event + - name: timer + - name: signal + - name: eventpolls + - name: other + - name: app_group.uptime + description: Apps uptime + unit: seconds + chart_type: line dimensions: - - name: a dimension per app group - - name: apps.pipes - description: Apps Open Pipes - unit: "open pipes" - chart_type: stacked + - name: uptime + - name: app_group.uptime_summary + description: Apps uptime summary + unit: seconds + chart_type: area dimensions: - - name: a dimension per app group + - name: min + - name: avg + - name: max - meta: plugin_name: apps.plugin module_name: groups @@ -289,160 +252,123 @@ modules: description: "" availability: [] scopes: - - name: global - description: "" - labels: [] + - name: user group + description: These metrics refer to the user group. + labels: + - name: user_group + description: The name of the user group. metrics: - - name: groups.cpu - description: User Groups CPU Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.cpu_user - description: User Groups CPU User Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.cpu_system - description: User Groups CPU System Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.cpu_guest - description: User Groups CPU Guest Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.mem - description: User Groups Real Memory (w/o shared) - unit: "MiB" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.rss - description: User Groups Resident Set Size (w/shared) - unit: "MiB" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.vmem - description: User Groups Virtual Memory Size - unit: "MiB" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.swap - description: User Groups Swap Memory - unit: "MiB" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.major_faults - description: User Groups Major Page Faults (swap read) - unit: "page faults/s" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.minor_faults - description: User Groups Page Faults (swap read) - unit: "page faults/s" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.preads - description: User Groups Disk Reads - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.pwrites - description: User Groups Disk Writes - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.lreads - description: User Groups Disk Logical Reads - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.lwrites - description: User Groups I/O Logical Writes - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.threads - description: User Groups Threads - unit: "threads" + - name: usergroup.cpu_utilization + description: User Groups CPU utilization (100% = 1 core) + unit: percentage chart_type: stacked dimensions: - - name: a dimension per user group - - name: groups.processes - description: User Groups Processes - unit: "processes" - chart_type: stacked - dimensions: - - name: a dimension per user group - - name: groups.voluntary_ctxt_switches - description: User Groups Voluntary Context Switches - unit: "processes" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: groups.involuntary_ctxt_switches - description: User Groups Involuntary Context Switches - unit: "processes" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: groups.uptime - description: User Groups Carried Over Uptime - unit: "seconds" + - name: user + - name: system + - name: usergroup.cpu_guest_utilization + description: User Groups CPU guest utilization (100% = 1 core) + unit: percentage chart_type: line dimensions: - - name: a dimension per user group - - name: groups.uptime_min - description: User Groups Minimum Uptime - unit: "seconds" + - name: guest + - name: usergroup.cpu_context_switches + description: User Groups CPU context switches + unit: switches/s + chart_type: stacked + dimensions: + - name: voluntary + - name: involuntary + - name: usergroup.mem_usage + description: User Groups memory RSS usage + unit: MiB + chart_type: area + dimensions: + - name: rss + - name: usergroup.mem_private_usage + description: User Groups memory usage without shared + unit: MiB + chart_type: area + dimensions: + - name: mem + - name: usergroup.vmem_usage + description: User Groups virtual memory size + unit: MiB chart_type: line dimensions: - - name: a dimension per user group - - name: groups.uptime_avg - description: User Groups Average Uptime - unit: "seconds" + - name: vmem + - name: usergroup.mem_page_faults + description: User Groups memory page faults + unit: pgfaults/s + chart_type: stacked + dimensions: + - name: minor + - name: major + - name: usergroup.swap_usage + description: User Groups swap usage + unit: MiB + chart_type: area + dimensions: + - name: swap + - name: usergroup.disk_physical_io + description: User Groups disk physical IO + unit: KiB/s + chart_type: area + dimensions: + - name: reads + - name: writes + - name: usergroup.disk_logical_io + description: User Groups disk logical IO + unit: KiB/s + chart_type: area + dimensions: + - name: reads + - name: writes + - name: usergroup.processes + description: User Groups processes + unit: processes chart_type: line dimensions: - - name: a dimension per user group - - name: groups.uptime_max - description: User Groups Maximum Uptime - unit: "seconds" + - name: processes + - name: usergroup.threads + description: User Groups threads + unit: threads chart_type: line dimensions: - - name: a dimension per user group - - name: groups.files - description: User Groups Open Files - unit: "open files" - chart_type: stacked + - name: threads + - name: usergroup.fds_open_limit + description: User Groups open file descriptors limit + unit: percentage + chart_type: line dimensions: - - name: a dimension per user group - - name: groups.sockets - description: User Groups Open Sockets - unit: "open sockets" - chart_type: stacked + - name: limit + - name: usergroup.fds_open + description: User Groups open file descriptors + unit: fds + chart_type: stacked + dimensions: + - name: files + - name: sockets + - name: pipes + - name: inotifies + - name: event + - name: timer + - name: signal + - name: eventpolls + - name: other + - name: usergroup.uptime + description: User Groups uptime + unit: seconds + chart_type: line dimensions: - - name: a dimension per user group - - name: groups.pipes - description: User Groups Open Pipes - unit: "open pipes" - chart_type: stacked + - name: uptime + - name: usergroup.uptime_summary + description: User Groups uptime summary + unit: seconds + chart_type: area dimensions: - - name: a dimension per user group + - name: min + - name: avg + - name: max - meta: plugin_name: apps.plugin module_name: users @@ -509,157 +435,120 @@ modules: description: "" availability: [] scopes: - - name: global - description: "" - labels: [] + - name: user + description: These metrics refer to the user. + labels: + - name: user + description: The name of the user. metrics: - - name: users.cpu - description: Users CPU Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.cpu_user - description: Users CPU User Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.cpu_system - description: Users CPU System Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.cpu_guest - description: Users CPU Guest Time (100% = 1 core) - unit: "percentage" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.mem - description: Users Real Memory (w/o shared) - unit: "MiB" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.rss - description: Users Resident Set Size (w/shared) - unit: "MiB" + - name: user.cpu_utilization + description: User CPU utilization (100% = 1 core) + unit: percentage chart_type: stacked dimensions: - - name: a dimension per user - - name: users.vmem - description: Users Virtual Memory Size - unit: "MiB" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.swap - description: Users Swap Memory - unit: "MiB" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.major_faults - description: Users Major Page Faults (swap read) - unit: "page faults/s" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.minor_faults - description: Users Page Faults (swap read) - unit: "page faults/s" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.preads - description: Users Disk Reads - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.pwrites - description: Users Disk Writes - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.lreads - description: Users Disk Logical Reads - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.lwrites - description: Users I/O Logical Writes - unit: "KiB/s" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.threads - description: Users Threads - unit: "threads" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.processes - description: Users Processes - unit: "processes" - chart_type: stacked - dimensions: - - name: a dimension per user - - name: users.voluntary_ctxt_switches - description: Users Voluntary Context Switches - unit: "processes" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: users.involuntary_ctxt_switches - description: Users Involuntary Context Switches - unit: "processes" - chart_type: stacked - dimensions: - - name: a dimension per app group - - name: users.uptime - description: Users Carried Over Uptime - unit: "seconds" + - name: user + - name: system + - name: user.cpu_guest_utilization + description: User CPU guest utilization (100% = 1 core) + unit: percentage chart_type: line dimensions: - - name: a dimension per user - - name: users.uptime_min - description: Users Minimum Uptime - unit: "seconds" + - name: guest + - name: user.cpu_context_switches + description: User CPU context switches + unit: switches/s + chart_type: stacked + dimensions: + - name: voluntary + - name: involuntary + - name: user.mem_usage + description: User memory RSS usage + unit: MiB + chart_type: area + dimensions: + - name: rss + - name: user.mem_private_usage + description: User memory usage without shared + unit: MiB + chart_type: area + dimensions: + - name: mem + - name: user.vmem_usage + description: User virtual memory size + unit: MiB chart_type: line dimensions: - - name: a dimension per user - - name: users.uptime_avg - description: Users Average Uptime - unit: "seconds" + - name: vmem + - name: user.mem_page_faults + description: User memory page faults + unit: pgfaults/s + chart_type: stacked + dimensions: + - name: minor + - name: major + - name: user.swap_usage + description: User swap usage + unit: MiB + chart_type: area + dimensions: + - name: swap + - name: user.disk_physical_io + description: User disk physical IO + unit: KiB/s + chart_type: area + dimensions: + - name: reads + - name: writes + - name: user.disk_logical_io + description: User disk logical IO + unit: KiB/s + chart_type: area + dimensions: + - name: reads + - name: writes + - name: user.processes + description: User processes + unit: processes chart_type: line dimensions: - - name: a dimension per user - - name: users.uptime_max - description: Users Maximum Uptime - unit: "seconds" + - name: processes + - name: user.threads + description: User threads + unit: threads chart_type: line dimensions: - - name: a dimension per user - - name: users.files - description: Users Open Files - unit: "open files" - chart_type: stacked + - name: threads + - name: user.fds_open_limit + description: User open file descriptors limit + unit: percentage + chart_type: line dimensions: - - name: a dimension per user - - name: users.sockets - description: Users Open Sockets - unit: "open sockets" - chart_type: stacked + - name: limit + - name: user.fds_open + description: User open file descriptors + unit: fds + chart_type: stacked + dimensions: + - name: files + - name: sockets + - name: pipes + - name: inotifies + - name: event + - name: timer + - name: signal + - name: eventpolls + - name: other + - name: user.uptime + description: User uptime + unit: seconds + chart_type: line dimensions: - - name: a dimension per user - - name: users.pipes - description: Users Open Pipes - unit: "open pipes" - chart_type: stacked + - name: uptime + - name: user.uptime_summary + description: User uptime summary + unit: seconds + chart_type: area dimensions: - - name: a dimension per user + - name: min + - name: avg + - name: max -- cgit v1.2.3