diff options
author | Daniel Baumann <daniel.baumann@progress-linux.org> | 2022-08-12 07:26:17 +0000 |
---|---|---|
committer | Daniel Baumann <daniel.baumann@progress-linux.org> | 2022-08-12 07:26:17 +0000 |
commit | 7877a98bd9c00db5e81dd2f8c734cba2bab20be7 (patch) | |
tree | d18b767250f7c7ced9b8abe2ece784ac1fe24d3e /collectors/diskspace.plugin | |
parent | Releasing debian version 1.35.1-2. (diff) | |
download | netdata-7877a98bd9c00db5e81dd2f8c734cba2bab20be7.tar.xz netdata-7877a98bd9c00db5e81dd2f8c734cba2bab20be7.zip |
Merging upstream version 1.36.0.
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to 'collectors/diskspace.plugin')
-rw-r--r-- | collectors/diskspace.plugin/plugin_diskspace.c | 501 |
1 files changed, 370 insertions, 131 deletions
diff --git a/collectors/diskspace.plugin/plugin_diskspace.c b/collectors/diskspace.plugin/plugin_diskspace.c index 663bb82fc..5bdf8bc61 100644 --- a/collectors/diskspace.plugin/plugin_diskspace.c +++ b/collectors/diskspace.plugin/plugin_diskspace.c @@ -3,11 +3,17 @@ #include "../proc.plugin/plugin_proc.h" #define PLUGIN_DISKSPACE_NAME "diskspace.plugin" +#define THREAD_DISKSPACE_SLOW_NAME "PLUGIN[diskspace slow]" #define DEFAULT_EXCLUDED_PATHS "/proc/* /sys/* /var/run/user/* /run/user/* /snap/* /var/lib/docker/*" #define DEFAULT_EXCLUDED_FILESYSTEMS "*gvfs *gluster* *s3fs *ipfs *davfs2 *httpfs *sshfs *gdfs *moosefs fusectl autofs" #define CONFIG_SECTION_DISKSPACE "plugin:proc:diskspace" +#define MAX_STAT_USEC 10000LU +#define SLOW_UPDATE_EVERY 5 + +static netdata_thread_t *diskspace_slow_thread = NULL; + static struct mountinfo *disk_mountinfo_root = NULL; static int check_for_new_mountpoints_every = 15; static int cleanup_mount_points = 1; @@ -34,6 +40,9 @@ struct mount_point_metadata { int do_inodes; int shown_error; int updated; + int slow; + + DICTIONARY *chart_labels; size_t collected; // the number of times this has been collected @@ -52,13 +61,15 @@ static DICTIONARY *dict_mountpoints = NULL; #define rrdset_obsolete_and_pointer_null(st) do { if(st) { rrdset_is_obsolete(st); (st) = NULL; } } while(st) -int mount_point_cleanup(const char *name, void *entry, void *data) { +int mount_point_cleanup(const char *name, void *entry, int slow) { (void)name; - (void)data; - + struct mount_point_metadata *mp = (struct mount_point_metadata *)entry; if(!mp) return 0; + if (slow != mp->slow) + return 0; + if(likely(mp->updated)) { mp->updated = 0; return 0; @@ -84,34 +95,219 @@ int mount_point_cleanup(const char *name, void *entry, void *data) { return 0; } -// for the full list of protected mount points look at -// https://github.com/systemd/systemd/blob/1eb3ef78b4df28a9e9f464714208f2682f957e36/src/core/namespace.c#L142-L149 -// https://github.com/systemd/systemd/blob/1eb3ef78b4df28a9e9f464714208f2682f957e36/src/core/namespace.c#L180-L194 -static const char *systemd_protected_mount_points[] = { - "/home", - "/root", - "/usr", - "/boot", - "/efi", - "/etc", - NULL +int mount_point_cleanup_cb(const char *name, void *entry, void *data) { + UNUSED(data); + + return mount_point_cleanup(name, (struct mount_point_metadata *)entry, 0); +} + +// a copy of basic mountinfo fields +struct basic_mountinfo { + char *persistent_id; + char *root; + char *mount_point; + char *filesystem; + + struct basic_mountinfo *next; }; -int mount_point_is_protected(char *mount_point) +static struct basic_mountinfo *slow_mountinfo_tmp_root = NULL; +static netdata_mutex_t slow_mountinfo_mutex; + +static struct basic_mountinfo *basic_mountinfo_create_and_copy(struct mountinfo* mi) { - for (size_t i = 0; systemd_protected_mount_points[i] != NULL; i++) - if (!strcmp(mount_point, systemd_protected_mount_points[i])) - return 1; + struct basic_mountinfo *bmi = callocz(1, sizeof(struct basic_mountinfo)); + + if (mi) { + bmi->persistent_id = strdupz(mi->persistent_id); + bmi->root = strdupz(mi->root); + bmi->mount_point = strdupz(mi->mount_point); + bmi->filesystem = strdupz(mi->filesystem); + } - return 0; + return bmi; } -static inline void do_disk_space_stats(struct mountinfo *mi, int update_every) { +static void add_basic_mountinfo(struct basic_mountinfo **root, struct mountinfo *mi) +{ + if (!root) + return; + + struct basic_mountinfo *bmi = basic_mountinfo_create_and_copy(mi); + + bmi->next = *root; + *root = bmi; +}; + +static void free_basic_mountinfo(struct basic_mountinfo *bmi) +{ + if (bmi) { + freez(bmi->persistent_id); + freez(bmi->root); + freez(bmi->mount_point); + freez(bmi->filesystem); + + freez(bmi); + } +}; + +static void free_basic_mountinfo_list(struct basic_mountinfo *root) +{ + struct basic_mountinfo *bmi = root, *next; + + while (bmi) { + next = bmi->next; + free_basic_mountinfo(bmi); + bmi = next; + } +} + +static void calculate_values_and_show_charts( + struct basic_mountinfo *mi, + struct mount_point_metadata *m, + struct statvfs *buff_statvfs, + int update_every) +{ const char *family = mi->mount_point; const char *disk = mi->persistent_id; + // logic found at get_fs_usage() in coreutils + unsigned long bsize = (buff_statvfs->f_frsize) ? buff_statvfs->f_frsize : buff_statvfs->f_bsize; + + fsblkcnt_t bavail = buff_statvfs->f_bavail; + fsblkcnt_t btotal = buff_statvfs->f_blocks; + fsblkcnt_t bavail_root = buff_statvfs->f_bfree; + fsblkcnt_t breserved_root = bavail_root - bavail; + fsblkcnt_t bused = likely(btotal >= bavail_root) ? btotal - bavail_root : bavail_root - btotal; + +#ifdef NETDATA_INTERNAL_CHECKS + if(unlikely(btotal != bavail + breserved_root + bused)) + error("DISKSPACE: disk block statistics for '%s' (disk '%s') do not sum up: total = %llu, available = %llu, reserved = %llu, used = %llu", mi->mount_point, disk, (unsigned long long)btotal, (unsigned long long)bavail, (unsigned long long)breserved_root, (unsigned long long)bused); +#endif + + // -------------------------------------------------------------------------- + + fsfilcnt_t favail = buff_statvfs->f_favail; + fsfilcnt_t ftotal = buff_statvfs->f_files; + fsfilcnt_t favail_root = buff_statvfs->f_ffree; + fsfilcnt_t freserved_root = favail_root - favail; + fsfilcnt_t fused = ftotal - favail_root; + + if(m->do_inodes == CONFIG_BOOLEAN_AUTO && favail == (fsfilcnt_t)-1) { + // this file system does not support inodes reporting + // eg. cephfs + m->do_inodes = CONFIG_BOOLEAN_NO; + } + +#ifdef NETDATA_INTERNAL_CHECKS + if(unlikely(btotal != bavail + breserved_root + bused)) + error("DISKSPACE: disk inode statistics for '%s' (disk '%s') do not sum up: total = %llu, available = %llu, reserved = %llu, used = %llu", mi->mount_point, disk, (unsigned long long)ftotal, (unsigned long long)favail, (unsigned long long)freserved_root, (unsigned long long)fused); +#endif + + // -------------------------------------------------------------------------- + + int rendered = 0; + + if(m->do_space == CONFIG_BOOLEAN_YES || (m->do_space == CONFIG_BOOLEAN_AUTO && + (bavail || breserved_root || bused || + netdata_zero_metrics_enabled == CONFIG_BOOLEAN_YES))) { + if(unlikely(!m->st_space) || m->st_space->update_every != update_every) { + m->do_space = CONFIG_BOOLEAN_YES; + m->st_space = rrdset_find_active_bytype_localhost("disk_space", disk); + if(unlikely(!m->st_space || m->st_space->update_every != update_every)) { + char title[4096 + 1]; + snprintfz(title, 4096, "Disk Space Usage"); + m->st_space = rrdset_create_localhost( + "disk_space" + , disk + , NULL + , family + , "disk.space" + , title + , "GiB" + , PLUGIN_DISKSPACE_NAME + , NULL + , NETDATA_CHART_PRIO_DISKSPACE_SPACE + , update_every + , RRDSET_TYPE_STACKED + ); + } + + rrdset_update_rrdlabels(m->st_space, m->chart_labels); + + m->rd_space_avail = rrddim_add(m->st_space, "avail", NULL, (collected_number)bsize, 1024 * 1024 * 1024, RRD_ALGORITHM_ABSOLUTE); + m->rd_space_used = rrddim_add(m->st_space, "used", NULL, (collected_number)bsize, 1024 * 1024 * 1024, RRD_ALGORITHM_ABSOLUTE); + m->rd_space_reserved = rrddim_add(m->st_space, "reserved_for_root", "reserved for root", (collected_number)bsize, 1024 * 1024 * 1024, RRD_ALGORITHM_ABSOLUTE); + } + else + rrdset_next(m->st_space); + + rrddim_set_by_pointer(m->st_space, m->rd_space_avail, (collected_number)bavail); + rrddim_set_by_pointer(m->st_space, m->rd_space_used, (collected_number)bused); + rrddim_set_by_pointer(m->st_space, m->rd_space_reserved, (collected_number)breserved_root); + rrdset_done(m->st_space); + + rendered++; + } + + // -------------------------------------------------------------------------- + + if(m->do_inodes == CONFIG_BOOLEAN_YES || (m->do_inodes == CONFIG_BOOLEAN_AUTO && + (favail || freserved_root || fused || + netdata_zero_metrics_enabled == CONFIG_BOOLEAN_YES))) { + if(unlikely(!m->st_inodes) || m->st_inodes->update_every != update_every) { + m->do_inodes = CONFIG_BOOLEAN_YES; + m->st_inodes = rrdset_find_active_bytype_localhost("disk_inodes", disk); + if(unlikely(!m->st_inodes) || m->st_inodes->update_every != update_every) { + char title[4096 + 1]; + snprintfz(title, 4096, "Disk Files (inodes) Usage"); + m->st_inodes = rrdset_create_localhost( + "disk_inodes" + , disk + , NULL + , family + , "disk.inodes" + , title + , "inodes" + , PLUGIN_DISKSPACE_NAME + , NULL + , NETDATA_CHART_PRIO_DISKSPACE_INODES + , update_every + , RRDSET_TYPE_STACKED + ); + } + + rrdset_update_rrdlabels(m->st_inodes, m->chart_labels); + + m->rd_inodes_avail = rrddim_add(m->st_inodes, "avail", NULL, 1, 1, RRD_ALGORITHM_ABSOLUTE); + m->rd_inodes_used = rrddim_add(m->st_inodes, "used", NULL, 1, 1, RRD_ALGORITHM_ABSOLUTE); + m->rd_inodes_reserved = rrddim_add(m->st_inodes, "reserved_for_root", "reserved for root", 1, 1, RRD_ALGORITHM_ABSOLUTE); + } + else + rrdset_next(m->st_inodes); + + rrddim_set_by_pointer(m->st_inodes, m->rd_inodes_avail, (collected_number)favail); + rrddim_set_by_pointer(m->st_inodes, m->rd_inodes_used, (collected_number)fused); + rrddim_set_by_pointer(m->st_inodes, m->rd_inodes_reserved, (collected_number)freserved_root); + rrdset_done(m->st_inodes); + + rendered++; + } + + // -------------------------------------------------------------------------- + + if(likely(rendered)) + m->collected++; +} + +static inline void do_disk_space_stats(struct mountinfo *mi, int update_every) { + const char *disk = mi->persistent_id; + static SIMPLE_PATTERN *excluded_mountpoints = NULL; static SIMPLE_PATTERN *excluded_filesystems = NULL; + + usec_t slow_timeout = MAX_STAT_USEC * update_every; + int do_space, do_inodes; if(unlikely(!dict_mountpoints)) { @@ -139,6 +335,7 @@ static inline void do_disk_space_stats(struct mountinfo *mi, int update_every) { struct mount_point_metadata *m = dictionary_get(dict_mountpoints, mi->mount_point); if(unlikely(!m)) { + int slow = 0; char var_name[4096 + 1]; snprintfz(var_name, 4096, "plugin:proc:diskspace:%s", mi->mount_point); @@ -158,7 +355,9 @@ static inline void do_disk_space_stats(struct mountinfo *mi, int update_every) { // check if the mount point is a directory #2407 // but only when it is enabled by default #4491 if(def_space != CONFIG_BOOLEAN_NO || def_inodes != CONFIG_BOOLEAN_NO) { + usec_t start_time = now_monotonic_high_precision_usec(); struct stat bs; + if(stat(mi->mount_point, &bs) == -1) { error("DISKSPACE: Cannot stat() mount point '%s' (disk '%s', filesystem '%s', root '%s')." , mi->mount_point @@ -181,6 +380,9 @@ static inline void do_disk_space_stats(struct mountinfo *mi, int update_every) { def_inodes = CONFIG_BOOLEAN_NO; } } + + if ((now_monotonic_high_precision_usec() - start_time) > slow_timeout) + slow = 1; } do_space = config_get_boolean_ondemand(var_name, "space usage", def_space); @@ -191,6 +393,7 @@ static inline void do_disk_space_stats(struct mountinfo *mi, int update_every) { .do_inodes = do_inodes, .shown_error = 0, .updated = 0, + .slow = 0, .collected = 0, @@ -205,7 +408,19 @@ static inline void do_disk_space_stats(struct mountinfo *mi, int update_every) { .rd_inodes_reserved = NULL }; + mp.chart_labels = rrdlabels_create(); + rrdlabels_add(mp.chart_labels, "mount_point", mi->mount_point, RRDLABEL_SRC_AUTO); + rrdlabels_add(mp.chart_labels, "filesystem", mi->filesystem, RRDLABEL_SRC_AUTO); + rrdlabels_add(mp.chart_labels, "mount_root", mi->root, RRDLABEL_SRC_AUTO); + m = dictionary_set(dict_mountpoints, mi->mount_point, &mp, sizeof(struct mount_point_metadata)); + + m->slow = slow; + } + + if (m->slow) { + add_basic_mountinfo(&slow_mountinfo_tmp_root, mi); + return; } m->updated = 1; @@ -215,13 +430,15 @@ static inline void do_disk_space_stats(struct mountinfo *mi, int update_every) { if (unlikely( mi->flags & MOUNTINFO_READONLY && - !mount_point_is_protected(mi->mount_point) && + !(mi->flags & MOUNTINFO_IS_IN_SYSD_PROTECTED_LIST) && !m->collected && m->do_space != CONFIG_BOOLEAN_YES && m->do_inodes != CONFIG_BOOLEAN_YES)) return; + usec_t start_time = now_monotonic_high_precision_usec(); struct statvfs buff_statvfs; + if (statvfs(mi->mount_point, &buff_statvfs) < 0) { if(!m->shown_error) { error("DISKSPACE: failed to statvfs() mount point '%s' (disk '%s', filesystem '%s', root '%s')" @@ -234,135 +451,133 @@ static inline void do_disk_space_stats(struct mountinfo *mi, int update_every) { } return; } - m->shown_error = 0; - // logic found at get_fs_usage() in coreutils - unsigned long bsize = (buff_statvfs.f_frsize) ? buff_statvfs.f_frsize : buff_statvfs.f_bsize; + if ((now_monotonic_high_precision_usec() - start_time) > slow_timeout) + m->slow = 1; - fsblkcnt_t bavail = buff_statvfs.f_bavail; - fsblkcnt_t btotal = buff_statvfs.f_blocks; - fsblkcnt_t bavail_root = buff_statvfs.f_bfree; - fsblkcnt_t breserved_root = bavail_root - bavail; - fsblkcnt_t bused; - if(likely(btotal >= bavail_root)) - bused = btotal - bavail_root; - else - bused = bavail_root - btotal; + m->shown_error = 0; -#ifdef NETDATA_INTERNAL_CHECKS - if(unlikely(btotal != bavail + breserved_root + bused)) - error("DISKSPACE: disk block statistics for '%s' (disk '%s') do not sum up: total = %llu, available = %llu, reserved = %llu, used = %llu", mi->mount_point, disk, (unsigned long long)btotal, (unsigned long long)bavail, (unsigned long long)breserved_root, (unsigned long long)bused); -#endif + struct basic_mountinfo bmi; + bmi.mount_point = mi->mount_point; + bmi.persistent_id = mi->persistent_id; + bmi.filesystem = mi->filesystem; + bmi.root = mi->root; - // -------------------------------------------------------------------------- + calculate_values_and_show_charts(&bmi, m, &buff_statvfs, update_every); +} - fsfilcnt_t favail = buff_statvfs.f_favail; - fsfilcnt_t ftotal = buff_statvfs.f_files; - fsfilcnt_t favail_root = buff_statvfs.f_ffree; - fsfilcnt_t freserved_root = favail_root - favail; - fsfilcnt_t fused = ftotal - favail_root; +static inline void do_slow_disk_space_stats(struct basic_mountinfo *mi, int update_every) { + struct mount_point_metadata *m = dictionary_get(dict_mountpoints, mi->mount_point); - if(m->do_inodes == CONFIG_BOOLEAN_AUTO && favail == (fsfilcnt_t)-1) { - // this file system does not support inodes reporting - // eg. cephfs - m->do_inodes = CONFIG_BOOLEAN_NO; + m->updated = 1; + + struct statvfs buff_statvfs; + if (statvfs(mi->mount_point, &buff_statvfs) < 0) { + if(!m->shown_error) { + error("DISKSPACE: failed to statvfs() mount point '%s' (disk '%s', filesystem '%s', root '%s')" + , mi->mount_point + , mi->persistent_id + , mi->filesystem?mi->filesystem:"" + , mi->root?mi->root:"" + ); + m->shown_error = 1; + } + return; } + m->shown_error = 0; -#ifdef NETDATA_INTERNAL_CHECKS - if(unlikely(btotal != bavail + breserved_root + bused)) - error("DISKSPACE: disk inode statistics for '%s' (disk '%s') do not sum up: total = %llu, available = %llu, reserved = %llu, used = %llu", mi->mount_point, disk, (unsigned long long)ftotal, (unsigned long long)favail, (unsigned long long)freserved_root, (unsigned long long)fused); -#endif + calculate_values_and_show_charts(mi, m, &buff_statvfs, update_every); +} - // -------------------------------------------------------------------------- +static void diskspace_slow_worker_cleanup(void *ptr) +{ + UNUSED(ptr); - int rendered = 0; + info("cleaning up..."); - if(m->do_space == CONFIG_BOOLEAN_YES || (m->do_space == CONFIG_BOOLEAN_AUTO && - (bavail || breserved_root || bused || - netdata_zero_metrics_enabled == CONFIG_BOOLEAN_YES))) { - if(unlikely(!m->st_space)) { - m->do_space = CONFIG_BOOLEAN_YES; - m->st_space = rrdset_find_active_bytype_localhost("disk_space", disk); - if(unlikely(!m->st_space)) { - char title[4096 + 1]; - snprintfz(title, 4096, "Disk Space Usage"); - m->st_space = rrdset_create_localhost( - "disk_space" - , disk - , NULL - , family - , "disk.space" - , title - , "GiB" - , PLUGIN_DISKSPACE_NAME - , NULL - , NETDATA_CHART_PRIO_DISKSPACE_SPACE - , update_every - , RRDSET_TYPE_STACKED - ); - } + worker_unregister(); +} - m->rd_space_avail = rrddim_add(m->st_space, "avail", NULL, (collected_number)bsize, 1024 * 1024 * 1024, RRD_ALGORITHM_ABSOLUTE); - m->rd_space_used = rrddim_add(m->st_space, "used", NULL, (collected_number)bsize, 1024 * 1024 * 1024, RRD_ALGORITHM_ABSOLUTE); - m->rd_space_reserved = rrddim_add(m->st_space, "reserved_for_root", "reserved for root", (collected_number)bsize, 1024 * 1024 * 1024, RRD_ALGORITHM_ABSOLUTE); - } - else - rrdset_next(m->st_space); +#define WORKER_JOB_SLOW_MOUNTPOINT 0 +#define WORKER_JOB_SLOW_CLEANUP 1 - rrddim_set_by_pointer(m->st_space, m->rd_space_avail, (collected_number)bavail); - rrddim_set_by_pointer(m->st_space, m->rd_space_used, (collected_number)bused); - rrddim_set_by_pointer(m->st_space, m->rd_space_reserved, (collected_number)breserved_root); - rrdset_done(m->st_space); +struct slow_worker_data { + netdata_thread_t *slow_thread; + int update_every; +}; - rendered++; - } +void *diskspace_slow_worker(void *ptr) +{ + struct slow_worker_data *data = (struct slow_worker_data *)ptr; + + worker_register("DISKSPACE_SLOW"); + worker_register_job_name(WORKER_JOB_SLOW_MOUNTPOINT, "mountpoint"); + worker_register_job_name(WORKER_JOB_SLOW_CLEANUP, "cleanup"); - // -------------------------------------------------------------------------- + struct basic_mountinfo *slow_mountinfo_root = NULL; - if(m->do_inodes == CONFIG_BOOLEAN_YES || (m->do_inodes == CONFIG_BOOLEAN_AUTO && - (favail || freserved_root || fused || - netdata_zero_metrics_enabled == CONFIG_BOOLEAN_YES))) { - if(unlikely(!m->st_inodes)) { - m->do_inodes = CONFIG_BOOLEAN_YES; - m->st_inodes = rrdset_find_active_bytype_localhost("disk_inodes", disk); - if(unlikely(!m->st_inodes)) { - char title[4096 + 1]; - snprintfz(title, 4096, "Disk Files (inodes) Usage"); - m->st_inodes = rrdset_create_localhost( - "disk_inodes" - , disk - , NULL - , family - , "disk.inodes" - , title - , "inodes" - , PLUGIN_DISKSPACE_NAME - , NULL - , NETDATA_CHART_PRIO_DISKSPACE_INODES - , update_every - , RRDSET_TYPE_STACKED - ); - } + int slow_update_every = data->update_every > SLOW_UPDATE_EVERY ? data->update_every : SLOW_UPDATE_EVERY; - m->rd_inodes_avail = rrddim_add(m->st_inodes, "avail", NULL, 1, 1, RRD_ALGORITHM_ABSOLUTE); - m->rd_inodes_used = rrddim_add(m->st_inodes, "used", NULL, 1, 1, RRD_ALGORITHM_ABSOLUTE); - m->rd_inodes_reserved = rrddim_add(m->st_inodes, "reserved_for_root", "reserved for root", 1, 1, RRD_ALGORITHM_ABSOLUTE); + netdata_thread_cleanup_push(diskspace_slow_worker_cleanup, data->slow_thread); + + usec_t step = slow_update_every * USEC_PER_SEC; + heartbeat_t hb; + heartbeat_init(&hb); + + while(!netdata_exit) { + worker_is_idle(); + heartbeat_next(&hb, step); + + usec_t start_time = now_monotonic_high_precision_usec(); + + if (!dict_mountpoints) + continue; + + if(unlikely(netdata_exit)) break; + + // -------------------------------------------------------------------------- + // disk space metrics + + worker_is_busy(WORKER_JOB_SLOW_MOUNTPOINT); + + netdata_mutex_lock(&slow_mountinfo_mutex); + free_basic_mountinfo_list(slow_mountinfo_root); + slow_mountinfo_root = slow_mountinfo_tmp_root; + slow_mountinfo_tmp_root = NULL; + netdata_mutex_unlock(&slow_mountinfo_mutex); + + struct basic_mountinfo *bmi; + for(bmi = slow_mountinfo_root; bmi; bmi = bmi->next) { + do_slow_disk_space_stats(bmi, slow_update_every); + + if(unlikely(netdata_exit)) break; } - else - rrdset_next(m->st_inodes); - rrddim_set_by_pointer(m->st_inodes, m->rd_inodes_avail, (collected_number)favail); - rrddim_set_by_pointer(m->st_inodes, m->rd_inodes_used, (collected_number)fused); - rrddim_set_by_pointer(m->st_inodes, m->rd_inodes_reserved, (collected_number)freserved_root); - rrdset_done(m->st_inodes); + if(unlikely(netdata_exit)) break; - rendered++; + worker_is_busy(WORKER_JOB_SLOW_CLEANUP); + + for(bmi = slow_mountinfo_root; bmi; bmi = bmi->next) { + struct mount_point_metadata *m = dictionary_get(dict_mountpoints, bmi->mount_point); + + if (m) + mount_point_cleanup(bmi->mount_point, m, 1); + } + + usec_t dt = now_monotonic_high_precision_usec() - start_time; + if (dt > step) { + slow_update_every = (dt / USEC_PER_SEC) * 3 / 2; + if (slow_update_every % SLOW_UPDATE_EVERY) + slow_update_every += SLOW_UPDATE_EVERY - slow_update_every % SLOW_UPDATE_EVERY; + step = slow_update_every * USEC_PER_SEC; + } } - // -------------------------------------------------------------------------- + netdata_thread_cleanup_pop(1); - if(likely(rendered)) - m->collected++; + free_basic_mountinfo_list(slow_mountinfo_root); + + return NULL; } static void diskspace_main_cleanup(void *ptr) { @@ -373,6 +588,13 @@ static void diskspace_main_cleanup(void *ptr) { info("cleaning up..."); + if (diskspace_slow_thread) { + netdata_thread_join(*diskspace_slow_thread, NULL); + freez(diskspace_slow_thread); + } + + free_basic_mountinfo_list(slow_mountinfo_tmp_root); + static_thread->enabled = NETDATA_MAIN_THREAD_EXITED; } @@ -402,6 +624,19 @@ void *diskspace_main(void *ptr) { if(check_for_new_mountpoints_every < update_every) check_for_new_mountpoints_every = update_every; + netdata_mutex_init(&slow_mountinfo_mutex); + + diskspace_slow_thread = mallocz(sizeof(netdata_thread_t)); + + struct slow_worker_data slow_worker_data = {.slow_thread = diskspace_slow_thread, .update_every = update_every}; + + netdata_thread_create( + diskspace_slow_thread, + THREAD_DISKSPACE_SLOW_NAME, + NETDATA_THREAD_OPTION_JOINABLE, + diskspace_slow_worker, + &slow_worker_data); + usec_t step = update_every * USEC_PER_SEC; heartbeat_t hb; heartbeat_init(&hb); @@ -411,7 +646,6 @@ void *diskspace_main(void *ptr) { if(unlikely(netdata_exit)) break; - // -------------------------------------------------------------------------- // this is smart enough not to reload it every time @@ -421,26 +655,31 @@ void *diskspace_main(void *ptr) { // -------------------------------------------------------------------------- // disk space metrics + netdata_mutex_lock(&slow_mountinfo_mutex); + free_basic_mountinfo_list(slow_mountinfo_tmp_root); + slow_mountinfo_tmp_root = NULL; + struct mountinfo *mi; for(mi = disk_mountinfo_root; mi; mi = mi->next) { - if(unlikely(mi->flags & (MOUNTINFO_IS_DUMMY | MOUNTINFO_IS_BIND))) continue; // exclude mounts made by ProtectHome and ProtectSystem systemd hardening options - if(mi->flags & MOUNTINFO_READONLY && !strcmp(mi->root, mi->mount_point)) + // https://github.com/netdata/netdata/issues/11498#issuecomment-950982878 + if(mi->flags & MOUNTINFO_READONLY && mi->flags & MOUNTINFO_IS_IN_SYSD_PROTECTED_LIST && !strcmp(mi->root, mi->mount_point)) continue; worker_is_busy(WORKER_JOB_MOUNTPOINT); do_disk_space_stats(mi, update_every); if(unlikely(netdata_exit)) break; } + netdata_mutex_unlock(&slow_mountinfo_mutex); if(unlikely(netdata_exit)) break; if(dict_mountpoints) { worker_is_busy(WORKER_JOB_CLEANUP); - dictionary_walkthrough_read(dict_mountpoints, mount_point_cleanup, NULL); + dictionary_walkthrough_read(dict_mountpoints, mount_point_cleanup_cb, NULL); } } |