Plone 5.2.2 Cluster eats up RAM and wonder why?

[jensens (Jens W. Klein)]

At least in my case I copied the wrong buildout config and instead of setting [instance] threads=1 I set zserver-threads=1 which was ignored and threads used the default 4. This resulted in 4 large ZODB cache pools in memory.

So each of my 8 instances used >2GB instead of ~500MB RAM... With 1 CPU assigned this degraded performance too.

[jaroel (Roel Bruggink)]

Seems like expected behavior?
4 threads * 1 connection = 4 connections == 4 caches

I think I remember the ZOBD doing connection pooling, but that might not be implemented correctly in the wsgi setup?

[jensens (Jens W. Klein)]

@jaroel

Seems like expected behavior?

Sure, all fine. My fault: zserver-threads is (as the name suggests) for Zserver. I just told the story so other fools like me are warned

There is no warning if zserver-threads is set, its just ignored.

[jensens (Jens W. Klein)]

@jaroel

I think I remember the ZOBD doing connection pooling, but that might not be implemented correctly in the wsgi setup?

All fine here. The behavior is the same here as in Zserver.

[jaroel (Roel Bruggink)]

okiedokie!

I read the topic title as "[My ]Plone 5.2.2 Cluster eats up RAM and [I ]wonder why?"