Hello,
This is a follow on from Brians's mail a while back about a possible
memory leak within dcache. After restarting dcache on my admin node
yesterday you can see a definite rise in memory use on fal-pygrid-20,
our admin node.
http://fal-pygrid-17.lancs.ac.uk:8123/ganglia/?r=day&c=LCG-ServiceNodes&h=fal-pygrid-20.lancs.ac.uk
The rise in memory use on it's own isn't a concern (it'll be one hell
of an application if it used no RAM whatsoever!) , it's the apparent
slow rise of memory use till all resources are apparently consumed
that concerns me. I'm still looking into this, but the dip and rise in
memory use correspond perfectly in time with my restart. I'm not sure
if this is dcache leaking memory or simply hogging system, resources,
so Brian and I thought it should be brought to people's attention in
case anyone sees similar phenomena (that's my word of the week).
As an addition I haven't noticed any particularly slow response on 20.
A top run on the admin node shows a lot of java processes taking up a
lot of memory, but no one particular process is consuming it all. The
admin node has a healthy 2GB of RAM.
We could of course be worrying other naught, but better safe then sorry.
Matt
|