Hi There,
we are using Zabbix 6.0.9 and have recently started to use/test the "new" MSA 2040 Templates.
> https://www.zabbix.com/integrations/hpe_san
We have not yet upgraded to 6.0.10 which contains some fixes for preprocessing (load):
System our Proxy is running on has 16GB of Ram assigned, 4GB are supposed to be for its Database (Maria), 8GB are usually used by Proxy Processes.
Proxy has 2050 Hosts assigned and has a required performance of 2800 vps, its monitoring 235k items.
Picture says more than thousand words - System Memory usually looks like this:
After the MSA Template was assigned to a single Host/Storage it went like this:
You can see it started to gradually built up after the Template was assigned (around 17:15).
Around 18:37 the oom killer reaped our database (of course
) and the Proxy was basically dead.
Now the thing is, ive looked at the Template and the only thing that special (at least for me) is the Script part.
To cross check things we have disabled the Host using the MSA Template and memory usage somewhat has stabilized:
One could argue that the System was already tight on resources but i suspect that there might be leaked memory from Javascript execution used in the MSA Template.
I have also looked at the Zabbix Poller Memory consumption and it seems the 10 largest processes use average ~170MB of RES from which 122MB are Shared.
Are there known Problems in regards of Javascript execution leaking Memory or does it indeed just increase the Poller Memory footprint.
If it increases its footprint - does anyone have numbers so we could correctly resize the System?
Best Regards,
medl
we are using Zabbix 6.0.9 and have recently started to use/test the "new" MSA 2040 Templates.
> https://www.zabbix.com/integrations/hpe_san
We have not yet upgraded to 6.0.10 which contains some fixes for preprocessing (load):
| ZBX-21580 | Optimized HPE MSA 2060 Storage by HTTP template to spread the load across preprocessing workers |
| ZBX-21579 | Optimized HPE MSA 2040 Storage by HTTP template to spread the load across preprocessing workers |
Proxy has 2050 Hosts assigned and has a required performance of 2800 vps, its monitoring 235k items.
Picture says more than thousand words - System Memory usually looks like this:
After the MSA Template was assigned to a single Host/Storage it went like this:
You can see it started to gradually built up after the Template was assigned (around 17:15).
Around 18:37 the oom killer reaped our database (of course
) and the Proxy was basically dead.Now the thing is, ive looked at the Template and the only thing that special (at least for me) is the Script part.
To cross check things we have disabled the Host using the MSA Template and memory usage somewhat has stabilized:
One could argue that the System was already tight on resources but i suspect that there might be leaked memory from Javascript execution used in the MSA Template.
I have also looked at the Zabbix Poller Memory consumption and it seems the 10 largest processes use average ~170MB of RES from which 122MB are Shared.
Are there known Problems in regards of Javascript execution leaking Memory or does it indeed just increase the Poller Memory footprint.
If it increases its footprint - does anyone have numbers so we could correctly resize the System?
Best Regards,
medl