The secondary MDC, uses all its memory over about 12-24 hours (depending on use) and slows to a crawl kicking all the NFS clients. I need to reboot it to return the SAN to service.
We solved this previously by hosting the Volume on the opposite head from the MDC that was running the SmallTree Card and hosting the NFS client connections.
Two MDCs, 1 promise Vtrak tray (16tbs) running 10.6.4, 12 GBs of RAM each.
Our volume is hosted by the other MDC, (the one that uses up it's RAM) is the NFS connection for all clients.
Cujo:(12GBs) Hosting Volume Primary MDC
Jack: (the one running out of RAM (12GBs)) SmallTree 4 Port Card NFS connections Secondary MDC
Both running 10.6.4
I am getting this in /Library/Filesystems/Xsan/data/SHARE/logcvlog on the MDC hosting the volume: (But not sure it's related.)
[0228 11:21:03.002755] 0x7fff70c14c20 (Debug) Node [106] [cujo-meta.domain.com:50130] connected. [0228 11:21:03.003257] 0x7fff70c14c20 (Debug) Active service - NSS ping from cujo-meta.domain.com:50129. [0228 11:21:03] 0x11eb81000 (Info) Node [106] [cujo-meta.domain.c:50130] Administrator Logout. [0228 11:21:03.014043] 0x7fff70c14c20 (Debug) Node [107] [cujo-meta.domain.com:50134] connected. [0228 11:21:03.014504] 0x7fff70c14c20 (Debug) Active service - NSS ping from cujo-meta.domain.com:50133. [0228 11:21:03] 0x11eb81000 (Info) Node [107] [cujo-meta.domain.c:50134] Administrator Logout.
actually, the server is not releasing Inactive Memory. I had to purge the inactive mem cache to release some of the RAM. Not sure why the system is not doing it.
10 GBs (out of 12) of RAM is inactive, that doesn't seem right to me.
