Adding upstream version 1.18.1.upstream/1.18.1

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

5 files changed, 194 insertions, 39 deletions

diff --git a/daemon/Makefile.am b/daemon/Makefile.am
index e020e517b..ee1b53d09 100644
--- a/daemon/Makefile.am
+++ b/daemon/Makefile.am
@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: GPL-3.0-or-later
 
 AUTOMAKE_OPTIONS = subdir-objects
-MAINTAINERCLEANFILES= $(srcdir)/Makefile.in
+MAINTAINERCLEANFILES = $(srcdir)/Makefile.in
 CLEANFILES = \
 	anonymous-statistics.sh \
 	$(NULL)
diff --git a/daemon/global_statistics.c b/daemon/global_statistics.c
index 53b7546f2..2bcc5c9fd 100644
--- a/daemon/global_statistics.c
+++ b/daemon/global_statistics.c
@@ -534,11 +534,30 @@ void global_statistics_charts(void) {
     // ----------------------------------------------------------------
 
 #ifdef ENABLE_DBENGINE
-    if (localhost->rrd_memory_mode == RRD_MEMORY_MODE_DBENGINE) {
-        unsigned long long stats_array[RRDENG_NR_STATS];
+    RRDHOST *host;
+    unsigned long long stats_array[RRDENG_NR_STATS] = {0};
+    unsigned long long local_stats_array[RRDENG_NR_STATS];
+    unsigned hosts_with_dbengine = 0, i;
+
+    rrd_rdlock();
+    rrdhost_foreach_read(host) {
+        if (host->rrd_memory_mode == RRD_MEMORY_MODE_DBENGINE) {
+            ++hosts_with_dbengine;
+            /* get localhost's DB engine's statistics */
+            rrdeng_get_33_statistics(host->rrdeng_ctx, local_stats_array);
+            for (i = 0 ; i < RRDENG_NR_STATS ; ++i) {
+                /* aggregate statistics across hosts */
+                stats_array[i] += local_stats_array[i];
+            }
+        }
+    }
+    rrd_unlock();
 
-        /* get localhost's DB engine's statistics */
-        rrdeng_get_33_statistics(localhost->rrdeng_ctx, stats_array);
+    if (hosts_with_dbengine) {
+        /* deduplicate global statistics by getting the ones from the last host */
+        stats_array[30] = local_stats_array[30];
+        stats_array[31] = local_stats_array[31];
+        stats_array[32] = local_stats_array[32];
 
         // ----------------------------------------------------------------
 
@@ -639,7 +658,7 @@ void global_statistics_charts(void) {
             static RRDSET *st_pg_cache_pages = NULL;
             static RRDDIM *rd_descriptors = NULL;
             static RRDDIM *rd_populated = NULL;
-            static RRDDIM *rd_commited = NULL;
+            static RRDDIM *rd_committed = NULL;
             static RRDDIM *rd_insertions = NULL;
             static RRDDIM *rd_deletions = NULL;
             static RRDDIM *rd_backfills = NULL;
@@ -663,7 +682,7 @@ void global_statistics_charts(void) {
 
                 rd_descriptors = rrddim_add(st_pg_cache_pages, "descriptors", NULL, 1, 1, RRD_ALGORITHM_ABSOLUTE);
                 rd_populated = rrddim_add(st_pg_cache_pages, "populated", NULL, 1, 1, RRD_ALGORITHM_ABSOLUTE);
-                rd_commited = rrddim_add(st_pg_cache_pages, "commited", NULL, 1, 1, RRD_ALGORITHM_ABSOLUTE);
+                rd_committed = rrddim_add(st_pg_cache_pages, "committed", NULL, 1, 1, RRD_ALGORITHM_ABSOLUTE);
                 rd_insertions = rrddim_add(st_pg_cache_pages, "insertions", NULL, 1, 1, RRD_ALGORITHM_INCREMENTAL);
                 rd_deletions = rrddim_add(st_pg_cache_pages, "deletions", NULL, -1, 1, RRD_ALGORITHM_INCREMENTAL);
                 rd_backfills = rrddim_add(st_pg_cache_pages, "backfills", NULL, 1, 1, RRD_ALGORITHM_INCREMENTAL);
@@ -674,7 +693,7 @@ void global_statistics_charts(void) {
 
             rrddim_set_by_pointer(st_pg_cache_pages, rd_descriptors, (collected_number)stats_array[27]);
             rrddim_set_by_pointer(st_pg_cache_pages, rd_populated, (collected_number)stats_array[3]);
-            rrddim_set_by_pointer(st_pg_cache_pages, rd_commited, (collected_number)stats_array[4]);
+            rrddim_set_by_pointer(st_pg_cache_pages, rd_committed, (collected_number)stats_array[4]);
             rrddim_set_by_pointer(st_pg_cache_pages, rd_insertions, (collected_number)stats_array[5]);
             rrddim_set_by_pointer(st_pg_cache_pages, rd_deletions, (collected_number)stats_array[6]);
             rrddim_set_by_pointer(st_pg_cache_pages, rd_backfills, (collected_number)stats_array[9]);
@@ -818,6 +837,63 @@ void global_statistics_charts(void) {
             rrddim_set_by_pointer(st_fd, rd_fd_max, (collected_number)rlimit_nofile.rlim_cur / 4);
             rrdset_done(st_fd);
         }
+
+        // ----------------------------------------------------------------
+
+        {
+            static RRDSET *st_ram_usage = NULL;
+            static RRDDIM *rd_cached = NULL;
+            static RRDDIM *rd_pinned = NULL;
+            static RRDDIM *rd_metadata = NULL;
+
+            collected_number cached_pages, pinned_pages, API_producers, populated_pages, metadata, pages_on_disk,
+            page_cache_descriptors;
+
+            if (unlikely(!st_ram_usage)) {
+                st_ram_usage = rrdset_create_localhost(
+                "netdata"
+                , "dbengine_ram"
+                , NULL
+                , "dbengine"
+                , NULL
+                , "NetData DB engine RAM usage"
+                , "MiB"
+                , "netdata"
+                , "stats"
+                , 130509
+                , localhost->rrd_update_every
+                , RRDSET_TYPE_STACKED
+                );
+
+                rd_cached = rrddim_add(st_ram_usage, "cache", NULL, 1, 256, RRD_ALGORITHM_ABSOLUTE);
+                rd_pinned = rrddim_add(st_ram_usage, "collectors", NULL, 1, 256, RRD_ALGORITHM_ABSOLUTE);
+                rd_metadata = rrddim_add(st_ram_usage, "metadata", NULL, 1, 1048576, RRD_ALGORITHM_ABSOLUTE);
+            }
+            else
+                rrdset_next(st_ram_usage);
+
+            API_producers = (collected_number)stats_array[0];
+            pages_on_disk = (collected_number)stats_array[2];
+            populated_pages = (collected_number)stats_array[3];
+            page_cache_descriptors = (collected_number)stats_array[27];
+
+            if (API_producers * 2 > populated_pages) {
+                pinned_pages = API_producers;
+            } else{
+                pinned_pages = API_producers * 2;
+            }
+            cached_pages = populated_pages - pinned_pages;
+
+            metadata = page_cache_descriptors * sizeof(struct page_cache_descr);
+            metadata += pages_on_disk * sizeof(struct rrdeng_page_descr);
+            /* This is an empirical estimation for Judy array indexing and extent structures */
+            metadata += pages_on_disk * 58;
+
+            rrddim_set_by_pointer(st_ram_usage, rd_cached, cached_pages);
+            rrddim_set_by_pointer(st_ram_usage, rd_pinned, pinned_pages);
+            rrddim_set_by_pointer(st_ram_usage, rd_metadata, metadata);
+            rrdset_done(st_ram_usage);
+        }
     }
 #endif
 
diff --git a/daemon/main.c b/daemon/main.c
index 4189ac7bd..0e56654db 100644
--- a/daemon/main.c
+++ b/daemon/main.c
@@ -146,46 +146,28 @@ void web_server_config_options(void) {
 }
 
 
-int killpid(pid_t pid, int signal)
-{
-    int ret = -1;
+// killpid kills pid with SIGTERM.
+int killpid(pid_t pid) {
+    int ret;
     debug(D_EXIT, "Request to kill pid %d", pid);
 
     errno = 0;
-    if(kill(pid, 0) == -1) {
+    ret = kill(pid, SIGTERM);
+    if (ret == -1) {
         switch(errno) {
             case ESRCH:
-                error("Request to kill pid %d, but it is not running.", pid);
-                break;
+                // We wanted the process to exit so just let the caller handle.
+                return ret;
 
             case EPERM:
-                error("Request to kill pid %d, but I do not have enough permissions.", pid);
+                error("Cannot kill pid %d, but I do not have enough permissions.", pid);
                 break;
 
             default:
-                error("Request to kill pid %d, but I received an error.", pid);
+                error("Cannot kill pid %d, but I received an error.", pid);
                 break;
         }
     }
-    else {
-        errno = 0;
-        ret = kill(pid, signal);
-        if(ret == -1) {
-            switch(errno) {
-                case ESRCH:
-                    error("Cannot kill pid %d, but it is not running.", pid);
-                    break;
-
-                case EPERM:
-                    error("Cannot kill pid %d, but I do not have enough permissions.", pid);
-                    break;
-
-                default:
-                    error("Cannot kill pid %d, but I received an error.", pid);
-                    break;
-            }
-        }
-    }
 
     return ret;
 }
diff --git a/daemon/main.h b/daemon/main.h
index 687155981..9d9f4ef0f 100644
--- a/daemon/main.h
+++ b/daemon/main.h
@@ -41,7 +41,7 @@ struct netdata_static_thread {
 };
 
 extern void cancel_main_threads(void);
-extern int killpid(pid_t pid, int signal);
+extern int killpid(pid_t pid);
 extern void netdata_cleanup_and_exit(int ret) NORETURN;
 extern void send_statistics(const char *action, const char *action_result, const char *action_data);
 
diff --git a/daemon/signals.c b/daemon/signals.c
index 71f271887..5378b04e5 100644
--- a/daemon/signals.c
+++ b/daemon/signals.c
@@ -2,6 +2,8 @@
 
 #include "common.h"
 
+static int reaper_enabled = 0;
+
 typedef enum signal_action {
     NETDATA_SIGNAL_END_OF_LIST,
     NETDATA_SIGNAL_IGNORE,
@@ -10,6 +12,7 @@ typedef enum signal_action {
     NETDATA_SIGNAL_LOG_ROTATE,
     NETDATA_SIGNAL_RELOAD_HEALTH,
     NETDATA_SIGNAL_FATAL,
+    NETDATA_SIGNAL_CHILD,
 } SIGNAL_ACTION;
 
 static struct {
@@ -26,6 +29,7 @@ static struct {
         { SIGUSR1, "SIGUSR1", 0, NETDATA_SIGNAL_SAVE_DATABASE },
         { SIGUSR2, "SIGUSR2", 0, NETDATA_SIGNAL_RELOAD_HEALTH },
         { SIGBUS,  "SIGBUS",  0, NETDATA_SIGNAL_FATAL         },
+        { SIGCHLD, "SIGCHLD", 0, NETDATA_SIGNAL_CHILD         },
 
         // terminator
         { 0,       "NONE",    0, NETDATA_SIGNAL_END_OF_LIST   }
@@ -42,7 +46,7 @@ static void signal_handler(int signo) {
                 char buffer[200 + 1];
                 snprintfz(buffer, 200, "\nSIGNAL HANLDER: received: %s. Oops! This is bad!\n", signals_waiting[i].name);
                 if(write(STDERR_FILENO, buffer, strlen(buffer)) == -1) {
-                    // nothing to do - we cannot write but there is no way to complaint about it
+                    // nothing to do - we cannot write but there is no way to complain about it
                     ;
                 }
             }
@@ -74,15 +78,33 @@ void signals_init(void) {
     struct sigaction sa;
     sa.sa_flags = 0;
 
+    // Enable process tracking / reaper if running as init (pid == 1).
+    // This prevents zombie processes when running in a container.
+    if (getpid() == 1) {
+        info("SIGNAL: Enabling reaper");
+        myp_init();
+        reaper_enabled = 1;
+    } else {
+        info("SIGNAL: Not enabling reaper");
+    }
+
     // ignore all signals while we run in a signal handler
     sigfillset(&sa.sa_mask);
 
     int i;
     for (i = 0; signals_waiting[i].action != NETDATA_SIGNAL_END_OF_LIST; i++) {
-        if(signals_waiting[i].action == NETDATA_SIGNAL_IGNORE)
+        switch (signals_waiting[i].action) {
+        case NETDATA_SIGNAL_IGNORE:
             sa.sa_handler = SIG_IGN;
-        else
+            break;
+        case NETDATA_SIGNAL_CHILD:
+            if (reaper_enabled == 0)
+                continue;
+            // FALLTHROUGH
+        default:
             sa.sa_handler = signal_handler;
+            break;
+        }
 
         if(sigaction(signals_waiting[i].signo, &sa, NULL) == -1)
             error("SIGNAL: Failed to change signal handler for: %s", signals_waiting[i].name);
@@ -100,6 +122,76 @@ void signals_reset(void) {
         if(sigaction(signals_waiting[i].signo, &sa, NULL) == -1)
             error("SIGNAL: Failed to reset signal handler for: %s", signals_waiting[i].name);
     }
+
+    if (reaper_enabled == 1)
+        myp_free();
+}
+
+// reap_child reaps the child identified by pid.
+static void reap_child(pid_t pid) {
+    siginfo_t i;
+
+    errno = 0;
+    debug(D_CHILDS, "SIGNAL: Reaping pid: %d...", pid);
+    if (waitid(P_PID, (id_t)pid, &i, WEXITED|WNOHANG) == -1) {
+        if (errno != ECHILD)
+            error("SIGNAL: Failed to wait for: %d", pid);
+        else
+            debug(D_CHILDS, "SIGNAL: Already reaped: %d", pid);
+        return;
+    } else if (i.si_pid == 0) {
+        // Process didn't exit, this shouldn't happen.
+        return;
+    }
+
+    switch (i.si_code) {
+    case CLD_EXITED:
+        debug(D_CHILDS, "SIGNAL: Child %d exited: %d", pid, i.si_status);
+        break;
+    case CLD_KILLED:
+        debug(D_CHILDS, "SIGNAL: Child %d killed by signal: %d", pid, i.si_status);
+        break;
+    case CLD_DUMPED:
+        debug(D_CHILDS, "SIGNAL: Child %d dumped core by signal: %d", pid, i.si_status);
+        break;
+    case CLD_STOPPED:
+        debug(D_CHILDS, "SIGNAL: Child %d stopped by signal: %d", pid, i.si_status);
+        break;
+    case CLD_TRAPPED:
+        debug(D_CHILDS, "SIGNAL: Child %d trapped by signal: %d", pid, i.si_status);
+        break;
+    case CLD_CONTINUED:
+        debug(D_CHILDS, "SIGNAL: Child %d continued by signal: %d", pid, i.si_status);
+        break;
+    default:
+        debug(D_CHILDS, "SIGNAL: Child %d gave us a SIGCHLD with code %d and status %d.", pid, i.si_code, i.si_status);
+    }
+}
+
+// reap_children reaps all pending children which are not managed by myp.
+static void reap_children() {
+    siginfo_t i;
+
+    while (1 == 1) {
+        // Identify which process caused the signal so we can determine
+        // if we need to reap a re-parented process.
+        i.si_pid = 0;
+        if (waitid(P_ALL, (id_t)0, &i, WEXITED|WNOHANG|WNOWAIT) == -1) {
+            if (errno != ECHILD) // This shouldn't happen with WNOHANG but does.
+                error("SIGNAL: Failed to wait");
+            return;
+        } else if (i.si_pid == 0) {
+            // No child exited.
+            return;
+        } else if (myp_reap(i.si_pid) == 0) {
+            // myp managed, sleep for a short time to avoid busy wait while
+            // this is handled by myp.
+            usleep(10000);
+        } else {
+            // Unknown process, likely a re-parented child, reap it.
+            reap_child(i.si_pid);
+        }
+    }
 }
 
 void signals_handle(void) {
@@ -157,6 +249,11 @@ void signals_handle(void) {
                             case NETDATA_SIGNAL_FATAL:
                                 fatal("SIGNAL: Received %s. netdata now exits.", name);
 
+                            case NETDATA_SIGNAL_CHILD:
+                                debug(D_CHILDS, "SIGNAL: Received %s. Reaping...", name);
+                                reap_children();
+                                break;
+
                             default:
                                 info("SIGNAL: Received %s. No signal handler configured. Ignoring it.", name);
                                 break;