ExportXMLWordPrintable

    • Type: Problem report
    • Resolution: Unresolved
    • Priority: Trivial
    • None
    • Affects Version/s: 8.0.0
    • Component/s: Server (S)
    • Environment:
      os : ol8.10
      containers based - alpine
      2 Zabbix server active-standby with HA between them
      6 proxies
      2 DBs - 3 node for each and one active the second is syncer

      Summary: Zabbix server 8.0.0 supervisor process crashes with signal 11 every few hours (HA failover each time)

      Version: Zabbix 8.0.0 (revision c81e276), official container image zabbix/zabbix-server-pgsql:alpine-8.0.0
      (Alpine 3.24.3). A thin layer adds only curl, net-tools, mii-tool (verified via "apk list -I" diff: no
      library changed), so the stock image is affected too.

      Environment:

      • Direct upgrade from 7.0.30 (stable for over a month before). PostgreSQL 16 (Patroni), DB upgrade
          completed without errors ("database upgrade fully completed").
      • Server HA: 2 nodes (active/standby), docker, host network mode. 6 proxies (sqlite3) on 8.0.0.
      • About 426 hosts, 23k items, 8k triggers. No CEP rules, no Services, no SLA configured.

      Problem:
      Every few hours the active node stops. Log (the first line is always the same):
        One child process died (PID:18,exitcode/signal:11). Exiting ...
        cannot send RTC notification to service "supervisor" ... Connection refused
        HA manager has been paused ... Zabbix Server stopped. Zabbix 8.0.0 (revision c81e276).
      The dead PID is the supervisor process ("server #1 started supervisor #1"). No "Got signal" line
      or backtrace is printed. On one node (before the line above) we also see: "cannot receive data from CEP service".
      Container exits with code 0, restart policy restarts it, and the other node becomes active.
      Shutdown sometimes hangs 60s: "zabbix_server [7]: timed out while waiting for child processes".

      Crash times (UTC): 2026-11-11 12:01:56, 21:09:38; 2026-11-12 00:52:52, 07:44:23 (more on Nov 13-15).
      Precursor (some, not all crashes): internal item zabbix[process,service manager,avg,busy] stays at 100%
      for about 8 minutes before the crash. It was flat 0 for the days before the upgrade (graph attached).

      Not the cause (checked): no OOM, DB failover (Patroni timeline unchanged), time sync (chrony offsets in ms).

      Questions:
      1. Is this a known issue in 8.0.0? Is there a fix or workaround?
      2. How can we get a useful backtrace/core dump from the container (no "Fatal information" is logged)?

      Attachments: server logs from both nodes, "zabbix_server -R diaginfo" output, item graph.

            Assignee:
            Zabbix Support Team
            Reporter:
            Ali HAmmoud
            Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

              Created:
              Updated: