Hi, we have about 150 agents (not zabbix agent - home grown agent emulating zabbix agent) pushing each thousands of metrics. Originally we cached data on a host for 5 minutes and then send them in one bulk. Unfortunately that led into problems with performance of zabbix server - most of the time server wasn't able accept such a large messages and we got just "connection reset by peer" at agent side.
I lowered interval to 30s and also significantly increased number of trappers (from default 5 to 50). I would be interested in any other suggestions we might use to improve stability of such setup.
I lowered interval to 30s and also significantly increased number of trappers (from default 5 to 50). I would be interested in any other suggestions we might use to improve stability of such setup.
Comment