DB get more and more request

[thomas315]

Hello,

I'm trying to upgrade our "good" old 0.8.8h cacti to 1.2.12.
This is a quite big DB (around 20k hosts with 250k DS).

The new cacti architecture is the following.
One VM for WebServer and boost (64G / 8CPU) with good IO (for rrd writing) (SSD)
One VM for DB (64G)
3 VM for remote poller (16G . 4 CPU each)
OS is Freebsd12.1

Everything is ok but after a while (generaly 8 or 12h hours), the amount of request on the main DB begins to increase, and the whole system get stuck if I don't truncate polller_output and poller_output_boost.

You'll find the following metrics below.

I've got 30k bad indexes, that i'm trying to delete, but I don't find any other ideas to deal with that issue.

[thomas315]

Hello,

I haven't finish my post. But question is :
* Have you ever had this issue ?
* How can I troubleshoot ? (logs seems to be ok).

Thanks

Thomas

[bmfmancini]

That is some odd behavior

try doing a showproccesslist in sql and see what the connections are
Are you using spine ?
Multiple pollers ?

[thomas315]

Hello,

I use spine & multipoller (main poller + 3 remote pollers).

I'll try show processlist

Thanks

Thomas

[bmfmancini]

Oh BTW what mysql template are you using ?

[thomas315]

Hum I'dont remember exactly, I guess i got it from the web, and adapt it a litle bit, I can post the XML if neeeded.

Thomas

[bmfmancini]

Yes please I have been looking for one