Hello,
we have experienced a big bug when running with the default StartPollers=5 on a 8-core server. The problem arised when
1) many (>10) servers were monitored
AND (
2) zabbix_server was started
OR
3) monitored nodes were re-enabled)
The problem lies in concurrent updates and reads from multiple processes in zbxdbhigh/db.c in DBget_maxid_num(), where an endless loop may occur.
There are 3 solutions:
1) put StartPollers=1 in zabbix_server.conf
2) apply patch 1 - a temporary, but quite working patch
3) apply patch 2 - a more advanced locking mechanism - not tested thoroughly yet
Platform used was Solaris10-i386 and Oracle10. We believe the problem may occur in different setups as well.
Thank you
Tom
we have experienced a big bug when running with the default StartPollers=5 on a 8-core server. The problem arised when
1) many (>10) servers were monitored
AND (
2) zabbix_server was started
OR
3) monitored nodes were re-enabled)
The problem lies in concurrent updates and reads from multiple processes in zbxdbhigh/db.c in DBget_maxid_num(), where an endless loop may occur.
There are 3 solutions:
1) put StartPollers=1 in zabbix_server.conf
2) apply patch 1 - a temporary, but quite working patch
3) apply patch 2 - a more advanced locking mechanism - not tested thoroughly yet
Platform used was Solaris10-i386 and Oracle10. We believe the problem may occur in different setups as well.
Thank you
Tom