Skip to content

System log flooded with victoria-metrics warnings while an alert is active #1958

Closed

Description

@Tbaile

Steps to reproduce

  • Have any alert go active on the firewall, e.g. a service crashing or a WAN going down in a multi-WAN setup
  • Keep it active for at least 10 minutes
  • Check /var/log/messages

Expected behavior

No warnings from victoria-metrics while alerts are being raised and resolved

Actual behavior

victoria-metrics logs a warning plus an info line every 5 minutes for every alert that is active or just resolved, filling the system log for as long as the alert lasts:

victoria-metrics[8008]: 2026-10-22T10:10:12.949Z#011warn#011VictoriaMetrics/app/vmselect/promql/rollup_result_cache.go:77#011resetting rollup result cache because the metric ALERTS{service="nginx",db="os-metrics",instance="instance1",alertgroup="services",alertname="ServiceDown",alertstate="firing",severity="critical"} (Timestamp=1792663500000, Value=1.000000) has a timestamp older than -search.cacheTimestampOffset=5m0s by 12.000s
victoria-metrics[8008]: 2026-10-22T10:10:12.949Z#011info#011VictoriaMetrics/app/vmselect/promql/rollup_result_cache.go:217#011rollupResult cache has been cleared

While there, alerts are evaluated on data up to 5 and a half minutes old, so they fire and resolve later than they should.

Components

ns-plug 1.4.1
victoria-metrics 1.146.0

Activity

  1. added this to the 8.8.1 milestone on Oct 22, 2026
  2. self-assigned this
    on Oct 22, 2026
  3. added theissue type on Oct 22, 2026
  4. moved this from ToDo 🕐 to In Progress 🛠 in NethSecurityon Oct 22, 2026
  5. Tbaile commented on Oct 22, 2026

    @Tbaile
    CollaboratorAuthor

    Latest dev image

    QA: have an up-to-date machine with an alert, check that every 5 minutes you get the log showed in the issue.
    Flash over the new image, check that the issue is gone (but alerts in the UI are still there)

  6. added
    testingPackages are available from testing repositories
    on Oct 22, 2026
  7. moved this from In Progress 🛠 to Testing in NethSecurityon Oct 22, 2026
  8. m-dilorenzi commented on Oct 22, 2026

    @m-dilorenzi
    Collaborator
    1. Crash cron service
    root@NethSec:~# for i in 1 2 3 4 5 6 7; do kill -9 $(pidof crond); sleep 7; done
    kill: usage: kill [-s sigspec | -n signum | -sigspec] pid | jobspec ... or kill -l [sigspec]
    root@NethSec:~# /usr/libexec/telegraf-services | grep -o '{[^}]*cron[^}]*}'
    {"service": "cron", "instance": "instance1", "running": 0, "pid": 0, "exit_code": 137}
    
    1. Check the alarm has been opened and the issue is reproducible
    Image
    root@NethSec:~# grep "rollup_result_cache.go" /var/log/messages
    Oct 22 16:20:16 NethSec victoria-metrics[6525]: 2026-10-22T16:20:16.137Z#011warn#011VictoriaMetrics/app/vmselect/promql/rollup_result_cache.go:77#011resetting rollup result cache because the metric ALERTS{service="cron",db="os-metrics",instance="instance1",alertgroup="services",alertname="ServiceDown",alertstate="pending",severity="critical"} (Timestamp=1792685700000, Value=1.000000) has a timestamp older than -search.cacheTimestampOffset=5m0s by 15.000s
    Oct 22 16:20:16 NethSec victoria-metrics[6525]: 2026-10-22T16:20:16.137Z#011info#011VictoriaMetrics/app/vmselect/promql/rollup_result_cache.go:217#011rollupResult cache has been cleared
    Oct 22 16:25:16 NethSec victoria-metrics[6525]: 2026-10-22T16:25:16.173Z#011warn#011VictoriaMetrics/app/vmselect/promql/rollup_result_cache.go:77#011resetting rollup result cache because the metric ALERTS{service="cron",db="os-metrics",instance="instance1",alertgroup="services",alertname="ServiceDown",alertstate="pending",severity="critical"} (Timestamp=1792686000000, Value=NaN) has a timestamp older than -search.cacheTimestampOffset=5m0s by 15.000s
    Oct 22 16:25:16 NethSec victoria-metrics[6525]: 2026-10-22T16:25:16.173Z#011info#011VictoriaMetrics/app/vmselect/promql/rollup_result_cache.go:217#011rollupResult cache has been cleared
    
    1. Flash with the new image
    Image
    1. Crash cron service again
    root@NethSec:~# for i in 1 2 3 4 5 6 7; do kill -9 $(pidof crond); sleep 7; done
    kill: usage: kill [-s sigspec | -n signum | -sigspec] pid | jobspec ... or kill -l [sigspec]
    root@NethSec:~# /usr/libexec/telegraf-services | grep -o '{[^}]*cron[^}]*}'
    {"service": "cron", "instance": "instance1", "running": 0, "pid": 0, "exit_code": 137}
    root@NethSec:~# 
    
    1. Check the alarm has been opened again but the issue is no more reproducible (logs are not found inside /var/log/messages)
    Image
    root@NethSec:~# grep "rollup_result_cache.go" /var/log/messages
    root@NethSec:~#
    

    Verified

  9. added
    verifiedAll test cases were verified successfully
    and removed
    testingPackages are available from testing repositories
    on Oct 22, 2026
  10. removed their assignment
    on Oct 22, 2026
  11. moved this from Testing to Verified in NethSecurityon Oct 22, 2026
  12. moved this from Verified to Done ✅ in NethSecurityon Oct 27, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    verifiedAll test cases were verified successfully

    Type

    Projects

    Milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions