Re: Avoid insert host/service macros into DB
Posted: Tue Jul 13, 2021 1:47 pm
If you have IO wait on your DB server (or your XI server) it will impact the performance globally, generally sar doesn't pick up the spikes that often from my findings. If you run the top command and look at the wa column, how high does it get on the XI server/DB server?
I will need to investigate the system on a remote session to see what can be done, please create a ticket for this and include a link back to this forum thread so we can get a remote session setup to debug further:
https://support.nagios.com/tickets/
There are a lot of unique-to-the-organization factors in that equation, total services/types of checks and how long they run/how often you're checking/storage speed/system specs/network speed and utilization/etc being some examples. I noticed from the ticket you also have a custom component that's interacting with the database as well, all those things add up. The general rule of thumb that I use is that if the DB is offloaded and the kernel message queue isn't processing fast enough, and there isn't any obvious issues with the system, moving it back local seems to help. But I usually only see that on 14K+ total check systems. I usually recommend large systems keep the databases local.
Correct, pointing multiple XI systems at the same DB in an Active/Active setup is not supported at this time.
I'm investigating the customvariablestatus table question and will update shortly.
EDIT: I do not see a way to turn that off in the source code:
https://github.com/NagiosEnterprises/nd ... ablestatus
I will need to investigate the system on a remote session to see what can be done, please create a ticket for this and include a link back to this forum thread so we can get a remote session setup to debug further:
https://support.nagios.com/tickets/
There are a lot of unique-to-the-organization factors in that equation, total services/types of checks and how long they run/how often you're checking/storage speed/system specs/network speed and utilization/etc being some examples. I noticed from the ticket you also have a custom component that's interacting with the database as well, all those things add up. The general rule of thumb that I use is that if the DB is offloaded and the kernel message queue isn't processing fast enough, and there isn't any obvious issues with the system, moving it back local seems to help. But I usually only see that on 14K+ total check systems. I usually recommend large systems keep the databases local.
Correct, pointing multiple XI systems at the same DB in an Active/Active setup is not supported at this time.
I'm investigating the customvariablestatus table question and will update shortly.
EDIT: I do not see a way to turn that off in the source code:
https://github.com/NagiosEnterprises/nd ... ablestatus
