Uploaded image for project: 'ZABBIX BUGS AND ISSUES'
  1. ZABBIX BUGS AND ISSUES
  2. ZBX-24742

multiple nodata cause history delivery delay

XMLWordPrintable

    • Icon: Problem report Problem report
    • Resolution: Unresolved
    • Icon: Trivial Trivial
    • None
    • None
    • None
    • None

       

      I got few zabbix trapper items that receive data every second.
      If no triggers created - all is OK and data successfully pushed to db without lag
      If I create nodata()=1 trigger for each one - all is OK and data successfully pushed to db without lag
      If I create group trigger that must should appear if all items in group have no data for 1 minute - all item in group almost stop to go to DB, history cache start growing, lag for 'last data' for those items is growing. All items not in group - delivered OK until history cache overloading, after that - losing data for all items at all (zabbix inner metrics also lost) until restart.

      So what is the problem and why it work in such way? How to make trigger for group of 50+ items if even 5 already cause lag history delivery

      Group trigger example:

      nodata(/5413181cc5e2/acc180decf0f,1m)=1
      and nodata(/5413181cc5e2/889e4b3d4225,1m)=1
      and nodata(/5413181cc5e2/710d27574123,1m)=1
      and nodata(/5413181cc5e2/6c93a2671fb4,1m)=1
      and nodata(/5413181cc5e2/7f6d81bc66ae,1m)=1 

      **

      Steps to reproduce:

      1. Zabbix 6.0.30 with 2cpu, PG13 with 2cpu
      2. Create 10 zabbix trapper items in one host
      3. For each add nodata()=1 trigger
      4. Push at each with 1 metrics per second
      5. See last data, history cache stat and history table - it's ok,last data with 0-2 sec lag, cache usage is low, history data for all items go to db very fast
      6. For 5 of items make group trigger as decribed above
      7. Wait few minutes
      8. See last data, history cache stat and history table again - 5 items that not in group trigger work same, but 5 items that in group now have lag in latest data and in db they appear with huge lag, also history cache start incresing
      9. wait few hours to fill full history cache (or use very small cache size to speed up) - overloaded trapper processors, no data sent to DB, no latest data in UI, no trigger actions even for items not in group trigger
      10. after restarting back us to step 7

      Result:


      No special logs messages or crash dumps

      Expected:
      Group trigger wit few nodata() on different items shouldn't lock work with used items within

      Or in documentation must be some notions to avoid such triggers and why it wont work

            tbross Tomass Janis Bross
            mosinnik Nick
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated: