Hi all,
I've just set up a new Zabbix server that will be replacing two different alarming servers, and it works well, except that my queue of items to be updated basically gets plugged up with SNMPv1 items, which is puzzling to me.
This new server is querying servers (running the Zabbix agent), doing discovery on a dozen subnets or so and monitoring a host of virtual circuits, as well as some environmental monitoring equipment that reports status via SNMP. The servers never show up in the queue, but some of the SNMPv1 items end up waiting for 1-2 minutes.
We are running Zabbix 2.4.0 on a CentOS 6.5 server (quad core Intel processor, 8 GB memory) using the PostgreSQL backend.
Zabbix logs show lines like the following:
I can perform SNMP queries on these same hosts from the Zabbix tools using tools like snmpget and snmpwalk, and the values I get in return are extremely fast. There is plenty of network bandwidth between the Zabbix server and SNMP hosts (which number about 64). I have the server querying around 36 Linux and Unix servers.
I have increased PostgreSQL shared_buffers to 512MB and checkpoint_segments to 32, which seems to make PostgreSQL more responsive. I have also set Zabbix StartPollers to 150, StartPollersUnreachable to 24, StartPingers to 200 and StartDiscoverers to 10 (we are doing discovery on about a dozen (/24) subnets). I also have CacheSize set to 128MB and StartDBSyncers to 10.
Required server performance, new values per second currently reads at 61.83.
Suggestions? Help is very appreciated!
I've just set up a new Zabbix server that will be replacing two different alarming servers, and it works well, except that my queue of items to be updated basically gets plugged up with SNMPv1 items, which is puzzling to me.
This new server is querying servers (running the Zabbix agent), doing discovery on a dozen subnets or so and monitoring a host of virtual circuits, as well as some environmental monitoring equipment that reports status via SNMP. The servers never show up in the queue, but some of the SNMPv1 items end up waiting for 1-2 minutes.
We are running Zabbix 2.4.0 on a CentOS 6.5 server (quad core Intel processor, 8 GB memory) using the PostgreSQL backend.
Zabbix logs show lines like the following:
SNMP agent item "sample.snmp.data.value" on host "Host Cabinet 1" failed: first network error, wait for 15 seconds
temporarily disabling SNMP agent checks on host "Host Cabinet 1": host unavailable
resuming SNMP agent checks on host "Host Cabinet 1": connection restored
I have increased PostgreSQL shared_buffers to 512MB and checkpoint_segments to 32, which seems to make PostgreSQL more responsive. I have also set Zabbix StartPollers to 150, StartPollersUnreachable to 24, StartPingers to 200 and StartDiscoverers to 10 (we are doing discovery on about a dozen (/24) subnets). I also have CacheSize set to 128MB and StartDBSyncers to 10.
Required server performance, new values per second currently reads at 61.83.
Suggestions? Help is very appreciated!
Comment