Ad Widget

Collapse

Zabbix 8.0.0 — no alerts are created at all: alert manager receives no events

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • zabbix_frm84
    Junior Member
    • Nov 2026
    • 2

    #1

    Zabbix 8.0.0 — no alerts are created at all: alert manager receives no events

    Title:
    Zabbix 8.0.0 — no alerts are created at all: alert manager receives no events (updated 2026-11-19)

    Hello everyone,

    Zabbix 8.0.0 on Ubuntu 24.04, upgraded from 7.4.14 on 2026-11-06.
    Since 2026-11-16 10:41:28 UTC, no alerts are created at all, for any action.
    The action log is empty and the alerts table has no new automatic row after alertid 1410.
    Events and problems are still created normally.

    What I have checked:
    - alert manager, escalator and alert syncer are alive, 0% busy.
    - At LogLevel 4, the alert manager only logs "select null" every 60 s and receives no event at all.
    - The escalator only runs process_escalations() and processes 0 escalations.
    - Media type enabled, user media active (severity 56), Admin is Super admin.
    - Actions enabled, conditions correct, no maintenance, no CEP rules, events not suppressed.
    - CHECK TABLE alerts OK; no lock in SHOW PROCESSLIST.
    - config_cache_reload and two service restarts: no change.
    - A brand-new action with no conditions produced nothing.
    - Another new action (host + severity >= 3) produced nothing for 4 real events.
    - The proxies (7.4.11 at the time) have since been migrated to 8.0.0: unchanged.
    - A manual Telegram media-type test works: the message is received. The media/token is therefore not the problem.
    - Database version:
    dbversionid 1, mandatory 8000000, optional 8000000

    Important new evidence:
    - Escalations ARE created, but alerts stays empty.
    Example:
    eventid 12323802
    actionid 11
    escalationid 444
    status 2
    No row in alerts for this eventid.
    - alert manager still reports "sent 0, failed 0".
    - SHOW ENGINE INNODB STATUS shows neither a foreign-key error nor a deadlock.
    So MySQL is not rejecting the insert: the alert manager does not even attempt it.

    Schema comparison:
    - I compared the live schema with the official 8.0 schema
    (/usr/share/zabbix/sql-scripts/mysql/server.sql.gz).
    - 230 tables in both, no missing table, no extra table.
    - No AUTO_INCREMENT expected by the official schema is missing.
    - The FK constraints on alerts (c_alerts_1..6) are present in the official schema.
    - The schema therefore does not appear to be incomplete.

    Recurring in the log at every start:
    "[event processor #10] thread stopped", then #9, #8...
    at roughly 5 min 14 s intervals, leaving only #1 and #2 alive.

    Questions:
    1. Why does the alert manager receive no events while events are still created in the database?
    2. Are the "event processor thread stopped" messages an abnormal termination, and could they be the missing link between event creation and the alert manager?
    3. Is this a known issue in Zabbix 8.0.0, or a corrupt internal state after the 7.4.14 -> 8.0.0 upgrade?
    4. What would be the safe way to repair the event -> escalation -> alert pipeline without restoring the official schema over the existing database?

    Full logs and a sanitized schema dump are available on request.
    Thanks in advance.
    Last edited by zabbix_frm84; 19-11-2026, 22:18.
  • zabbix_frm84
    Junior Member
    • Nov 2026
    • 2

    #2
    [SOLVED] Zabbix 8.0.0 — no alerts are created at all

    Hello everyone,

    Quick update to close this thread: this was NOT a Zabbix 8.0.0 bug, and the schema was fine. The root cause was a configuration change on our side.

    Root cause:
    The only media of the Admin user (Telegram) had been disabled on 2026-11-16 at 18:49 UTC during unrelated configuration work done via the API. The audit log shows it clearly:

    user.medias[1].active: ["update","1","0"]

    Note that in the "media" table, active = 0 means ENABLED and active = 1 means DISABLED. During troubleshooting, "active = 1" was misread as "active", which sent us in the wrong direction.

    Why it looked so confusing:
    - Escalations were created and ended in status 2 (SLEEP). This is normal: the action had a single step, the step was executed, and the escalation then waits for recovery.
    - When the operation finds no enabled media for the recipient, Zabbix silently creates no alert: no row in "alerts", nothing in the action log, and the alert manager has nothing to send ("sent 0, failed 0").
    - The media type "Test" button still worked, because it does not use the user's media.
    - The "[event processor #N] thread stopped" messages are unrelated: they are the normal scale-down of idle workers (from 10 to 2, about every 5 minutes), and they were already present when alerts were working.
    - The start time we had in mind (10:41 UTC) was only the time of the last alert sent, not the time of the change.

    Fix:
    Users > Users > Admin > Media tab > Edit > Enabled > Update (then Update again on the user form). The next new problem was notified on Telegram immediately. Problems that were already open are not re-notified, because their escalation step has already been executed.

    Lesson learned: when notifications stop, check the auditlog table first, for example:
    SELECT FROM_UNIXTIME(clock), username, ip, LEFT(details,250)
    FROM auditlog WHERE resourcetype = 0 ORDER BY clock DESC LIMIT 20;
    and remember that media.active = 0 means enabled.

    Thanks to everyone who read this, and sorry for the noise.

    Comment

    Working...