Hi -- I am currently in the process of "upgrading" from zabbix 1.8.X to zabbix 3.2.3. I am NOT copying the database, or doing any sort of schema upgrade process, this is a fresh install of zabbix server 3.2.3. The only thing I am doing is importing my templates from zabbix 1.8, which has gone well as far as I can tell with no errors, etc.
I have been using vfs.file.cksum to record checksums of mysql backup files successfully for years, without any substantial load. However, when I upgraded to zabbix SERVER 3.2.3, my vfs.file.cksum rule started causing high load, and the agent became mostly unresponsive (leaving large gaps in ALL my graphs, not just vfs.file.cksum graphs). I tried also upgrading my agent to 3.2.3, but the results are 100% identical. I have narrowed it down to vfs.file.cksum by disabling everything else, and turning it on and off.
* CentOS 6 with most recent updates (server and agent)
* My vfs.file.cksum item has an interval of every 2 hours on zabbix server (this has been the case for years), but the problem is constant regardless of whether it should be running or not.
* The file is about 3.5G
* Agent timeout is set for 5 seconds (in both 1.8 and 3.2 agent), and has been for years with 1.8 server and agent.
* linux "time" reports that running cksum or md5sum from the CLI takes about 30 seconds.
* I'm seeing this on more than one server. The server I've used as an example here, and done most of my testing on is completely idle, without any other load.
* I get the same results with both 1.8 and 3.2.3 agent's, the common thread being zabbix server 3.2.3.
Let me know if theres any other information you need to help figure this out. Thanks!
Here is a screenshot showing the zabbix agent being unresponsive, and also the high load. at 10:56, you can see where I disabled the cksum item via the zabbix UI. at 11:04, you can see where I manually run cksum from the CLI to verify it worked as expected.
Item Config:
I have been using vfs.file.cksum to record checksums of mysql backup files successfully for years, without any substantial load. However, when I upgraded to zabbix SERVER 3.2.3, my vfs.file.cksum rule started causing high load, and the agent became mostly unresponsive (leaving large gaps in ALL my graphs, not just vfs.file.cksum graphs). I tried also upgrading my agent to 3.2.3, but the results are 100% identical. I have narrowed it down to vfs.file.cksum by disabling everything else, and turning it on and off.
* CentOS 6 with most recent updates (server and agent)
* My vfs.file.cksum item has an interval of every 2 hours on zabbix server (this has been the case for years), but the problem is constant regardless of whether it should be running or not.
* The file is about 3.5G
* Agent timeout is set for 5 seconds (in both 1.8 and 3.2 agent), and has been for years with 1.8 server and agent.
* linux "time" reports that running cksum or md5sum from the CLI takes about 30 seconds.
* I'm seeing this on more than one server. The server I've used as an example here, and done most of my testing on is completely idle, without any other load.
* I get the same results with both 1.8 and 3.2.3 agent's, the common thread being zabbix server 3.2.3.
Let me know if theres any other information you need to help figure this out. Thanks!
Here is a screenshot showing the zabbix agent being unresponsive, and also the high load. at 10:56, you can see where I disabled the cksum item via the zabbix UI. at 11:04, you can see where I manually run cksum from the CLI to verify it worked as expected.
Item Config: