mod_fcgid - gah! Why does this config change throw the server over a cliff?

Webmin
1

Hi all,

I’m running latest VirtualMin on a fairly stock Centos 7 VPS. Apache is using fcgid.

Just recently, I upgraded the VPS from 1 core / 2Gb of RAM to 2 cores / 4Gb of RAM.

However, if I now change FcgidMaxProcesses from 60 to 80 (while leaving everything else unchanged), to try and use this new capacity, the server immediately goes over a cliff. Load average climbs and climbs - but it seems possibly to be an IO-bound problem, as CPU use and memory use remain low according to htop.

One of the things I hate about fcgid is that there are so many different but apparently interdependent config settings - and wildly different approaches for how to configure it, once you start looking at sample configs online. Before I start to post detailed config files and run into all those complications, I’m just wondering if anybody has seen this particular problem and knows how best to resolve it, or what other parameters to try tweaking?

Thanks,
Kona

2

I’ve just come across suggestions that the limiting factor for FcgidMaxProcesses is the communication between apache and the fcgid processes. An IPC bottleneck reaching saturation would certainly go some way to explaining the problem above. The hosting company suggests not going over 50 for FcgidMaxProcesses.

It’s now not clear to me whether one should make changes to the mod_fcgid config when improving the server hardware spec, or not. I’ve certainly seen suggestions that one should tweak it for different configurations, e.g. the sample configs here: https://wiki.mikejung.biz/Fcgid - but a hard ceiling on FcgidMaxProcesses seems to contradict this.

Any experiences, anyone?