memory leak in autofs-4.1.3-114 on FC3/i386?
Hi,
an updated FC3/i386 box with mail server functionality accesses O(1000) accounts over NFS (NFS server is a dated Tru64 Unixcluster).
The automount process consumes memory until we have to kill it, or the machine starts oom-killing us. Here is a graph:
http://www.physik.fu-berlin.de/~dreger/memusage.png
(We have also seen bad behaviour of rpc.idmapd in the past, and 200MB looks a bit too much for mapping 1000 accounts)
As a side note: Even though there are only 1000 accounts on this system, mounts are attempted on each forged incoming address, e.g. there is a multitude of failing mount requests, perhaps that's causing the large rpc.idmapd process space and automount's memory consumption?
Anything else we can provide for debugging? Thanks.
Signature · 1 line
Axel.Thimm at ATrpms.net
_______________________________________________
autofs mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/autofs
Ian Kent
maintainerQuoted 10 lines · Axel Thimm
Hi,
an updated FC3/i386 box with mail server functionality accesses O(1000) accounts over NFS (NFS server is a dated Tru64 Unixcluster).
The automount process consumes memory until we have to kill it, or the machine starts oom-killing us. Here is a graph:
Quoted 8 lines
(We have also seen bad behaviour of rpc.idmapd in the past, and 200MB looks a bit too much for mapping 1000 accounts)
As a side note: Even though there are only 1000 accounts on this system, mounts are attempted on each forged incoming address, e.g. there is a multitude of failing mount requests, perhaps that's causing the large rpc.idmapd process space and automount's memory consumption?
I'm not sure I understand what's going on here.
Can you provide a debug log please.
Quoted 2 lines
Anything else we can provide for debugging? Thanks.
Jeff Moyer has a 4.1.4 rpm on his people page. Can you give that a try please.
Ian