Extreme memory usage of Zabbix Poller processes (Proxy)
I'm having a memory issue with one of our Zabbix proxies.
We have 8 proxies which we try to max out at around 1000-1300 NVPS.
One proxy, zabbix04, responsible for about half of our Active Agents, uses all the memory it can get. Way more than our other proxies, while also processing way less NVPS.
It also tends to have a queue of 3000-5000 while the others are steadily sitting in the low hundreds.
It has similar configuration to another Active Agent proxy (zabbix03), which barely uses any memory.
Sample numbers:
NVPS (30min avg):
RAM usage (all proxies have 12GB RAM):
zabbix03 = green, zabbix04 = orange/yellow (colorblind..), zabbix05 = red
All proxies run in Docker.
Environment files look like this:
The mysqld processes on zabbix03 uses about 670MB memory, and the Zabbix poller processes about 43K.
Comparatively, on zabbix04 mysqld uses 565M but poller processes about 105M, a massive increase.
I don't really see anything out of the ordinary in the container logs.
Not really sure where to look to solve these issues.
Any help would be greatly appreciated
We have 8 proxies which we try to max out at around 1000-1300 NVPS.
One proxy, zabbix04, responsible for about half of our Active Agents, uses all the memory it can get. Way more than our other proxies, while also processing way less NVPS.
It also tends to have a queue of 3000-5000 while the others are steadily sitting in the low hundreds.
It has similar configuration to another Active Agent proxy (zabbix03), which barely uses any memory.
Sample numbers:
NVPS (30min avg):
RAM usage (all proxies have 12GB RAM):
zabbix03 = green, zabbix04 = orange/yellow (colorblind..), zabbix05 = red
All proxies run in Docker.
Environment files look like this:
Code:
ZBX_CONFIGFREQUENCY=300 ZBX_CACHESIZE=1G ZBX_HISTORYCACHESIZE=384M ZBX_HISTORYINDEXCACHESIZE=96M ZBX_STARTPOLLERS=150 ZBX_STARTPOLLERSUNREACHABLE=30 ZBX_STARTPINGERS=20 ZBX_STARTVMWARECOLLECTORS=5 ZBX_VMWAREFREQUENCY=120 ZBX_VMWAREPERFFREQUENCY=120 ZBX_VMWARECACHESIZE=32M ZBX_VMWARETIMEOUT=10 ZBX_ENABLE_SNMP_TRAPS=true
Comparatively, on zabbix04 mysqld uses 565M but poller processes about 105M, a massive increase.
I don't really see anything out of the ordinary in the container logs.
Not really sure where to look to solve these issues.
Any help would be greatly appreciated
Comment