-
Incident report
-
Resolution: Unsupported version
-
Major
-
None
-
1.8.11, 2.0.0rc2
When a trigger has few items in expression it can be processed by few history syncers.
This problem appears only on heavy loaded installations, when one syncer cannot process all updated items (1000 pcs) from history cache (in memory) at one time.
an example (DebugLevel=4):
eventid from_unixtime(e.clock) eventid from_unixtime(e1.clock) key_ functionid itemid triggerid function parameter triggerid expression descriptio
3026295 2012-03-10 15:29:18 3026216 2012-03-10 15:29:19 icmppingloss 702056 1155559 417340 last 0 417340 ({702056}>20)&({702057}=1) Packets loss >20%
3026295 2012-03-10 15:29:18 3026216 2012-03-10 15:29:19 icmpping 702057 1155558 417340 last 0 417340 ({702056}>20)&({702057}=1) Packets loss >20%
the same but with another trigger:
3026487 2012-03-10 15:29:33 3026449 2012-03-10 15:29:34 icmppingloss 391839 682174 247319 last 0 247319 ({391839}>20)&({391840}=1) Packets loss >20%
3026487 2012-03-10 15:29:33 3026449 2012-03-10 15:29:34 icmpping 391840 682173 247319 last 0 247319 ({391839}>20)&({391840}=1) Packets loss >20%
and records from zabbix_server debug log (for first example):
cat test.log | grep 3026216 17924:20120310:152921.888 End of DBget_nextid():3026216 table:'events' recid:'eventid' 17924:20120310:152921.888 In process_event() eventid:3026216 object:0 objectid:417340 value:0 value_changed:0 17924:20120310:152921.888 query [txnlev:1] [insert into events (eventid,source,object,objectid,clock,ns,value,value_changed) values (3026216,0,0,417340,1331371759,450565427,0,0)]
cat test.log | grep 3026295 17919:20120310:152922.397 In process_event() eventid:3026295 object:0 objectid:417340 value:0 value_changed:0 17919:20120310:152922.397 query [txnlev:1] [insert into events (eventid,source,object,objectid,clock,ns,value,value_changed) values (3026295,0,0,417340,1331371758,662994086,0,0)]
We see that triggerid=417340 processed by two history syncers (PIDs: 17924, 17919) with interval in 0,5 second.