Hello Darren,

Friday, June 16, 2006, 3:26:06 PM, you wrote:

DJM> What a blast from the past this is! Thats my error message, so
DJM> sorry for it not being helpful.  

DJM> We assumed it wouldn't get hit my many people and that it would
DJM> be quite a while before it was, looks like that time has come :-)

DJM> Did increasing the size of the dupreq cache help avoid the problem ?

I did increased it upto 8192 value (both cotsmaxdupreqs and
maxdupreqs) on the live system via mdb (I know it could mean some
performance hit). So far so good but I guess we have to wait another
few days to be sure.

DJM> Out of interest, how many clients, filesystems are involved
DJM> here?  This probably was initially found on a system with about
DJM> 15TB of data spread out over about 1500 file systems, it was a
DJM> 10CPU E10k that just did NFS (assuming I remember it correctly).

~50 filesystems on this server with about 40TB data, about ~15 clients.
About 1500-2200 active nfsd threads on nfs server.
Also noticed it on other servers.

What I noticed (and you in a crashdump from me) that system could be
literally overflown with these messages - sometimes syslogd has
problems to recover and is eating lot of memory (few hundred MBs).

Best regards,
 Robert                            mailto:rmilkowski at task.gda.pl

Reply via email to