We recently migrated a Windows file server to a vfiler running on a pair of FAS3160s (OnTAP 220.127.116.11P7). After performing the cutover, we noticed extremely poor login performance for our XP lab workstations that are using mandatory (read-only) roaming profiles. We were seeing errors such as this: [cifs.oplock.break.timeout:warning]: CIFS: An oplock break request to station aaa.bbb.ccc.ddd() for filer VFILER, share Profiles, file \ProfilePath\Application Data\blah\blahblah\blah.xml has timed out, and through Performance Advisor noticed that the 'other_latency' counter spiked extremely (>100,000 microseconds) high on the volume hosting the profiles qtree. Read/Write latency, however, was low/negligible.
Despite everything we've heard about oplocks *improving* performance, we disabled oplocks for the Profiles qtree, and the performance problems immediately ceased.
1) If anyone else has come across a similar problem and
2) What 'other_latency' really means through Performance Advisor
I have also had issues with NetApp filers replacing Windows servers with the same name when folder redirections and offline folders were already in use then I have had to reinitialise the offline file cache due to clients randomly disconnecting and reconnecting, and appearing to lose documents.
You are also running about a 2 year old release (albeit with some patches). You might want to check the fixed bugs list against a release of a bit newer vintage... You can find release comparisons for bugs on the NOW site... should be 500 or so fixed since 18.104.22.168 ...