Hello, I'm testing out Zabbix for monitoring our network devices and possibly switching from Nagios and I've been having a few issues the largest of which is the DB connections getting larger and larger for max used and it doesn't appear that they are closing out in a timely matter. Since I'm just testing I'm currently running on a smaller box but after adding less than ten hosts I'm already at 6GB of RAM utilization out of the 8 I allocated and I keep having to restart services because eventually the connection amount to the database gets too high (although I've currently resolved this by moving max_connections up to 5k).
So back to this particular issue, yesterday after I finally resolved the connections increasing incredibly fast (which was caused by timeouts to the device, then the retries would open new connections while not closing the previously made ones), there were about 100 connections about 8 hours later it was up to 300 now this morning I just checked (about 18 hours later) and it's up to 550. Our ADCs do have around 150 graphs each to monitor separate VIPs and Pools so I'm sure that's contributing, Zabbix just seems really inefficient on utilization. Why are these increasing so much over time? Is this typically expected? Would it have anything to do with running discovery rules?
I need to eventually add about 100 more SNMP hosts on the network alone and probably around 750 servers so I think I'm at least looking at setting up a separate DB server. What settings should be done with an initial install? Would rolling back to version 2.x from 3.x be recommended?
Any tips, suggestions or thoughts are greatly appreciated!
So back to this particular issue, yesterday after I finally resolved the connections increasing incredibly fast (which was caused by timeouts to the device, then the retries would open new connections while not closing the previously made ones), there were about 100 connections about 8 hours later it was up to 300 now this morning I just checked (about 18 hours later) and it's up to 550. Our ADCs do have around 150 graphs each to monitor separate VIPs and Pools so I'm sure that's contributing, Zabbix just seems really inefficient on utilization. Why are these increasing so much over time? Is this typically expected? Would it have anything to do with running discovery rules?
I need to eventually add about 100 more SNMP hosts on the network alone and probably around 750 servers so I think I'm at least looking at setting up a separate DB server. What settings should be done with an initial install? Would rolling back to version 2.x from 3.x be recommended?
Any tips, suggestions or thoughts are greatly appreciated!