On Wed, 2007-03-14 at 15:52 -0400, Stewart Dean wrote: > 1) Since my index files are stored in on a separate, non-quotaed, not > NFS filesystem, can I correctly assume this means I can ignore the > nmap_disable and lock_method parms?
Right. > 2) The docs say: > > The easiest way to get better performance without caching problems is > > to just make sure the user is always redirected to the same server. If > > you can manage to do that, then it's safe to enable attribute caching. > The INBOXes and folders live on the mail server where > a) DC is going to run and > b) they are exported to the three other machine. > Does this mean that I don't have to mess with actimeo? So Dovecot runs in one server, but the mails are exported to 3 other servers too? If the other servers are accessing the mails at the same time as Dovecot, there can be problems if you don't set actimeo=0. Although I guess the worst that can happen is some "Invalid cached mbox offset" errors in log file which are transparently fixed. > 3) Pardon my ignorance....Is actimeo set on the NFS server as an > attribute of exportation or on the NFS client on the imported FS...or both It's a NFS client setting. > 4) Is there any reason related to data integrity or performance to > prefer NFS V2 or 3 or 4? I don't think you should use NFSv2, but I don't know if NFSv4 is much better than NFSv3. Although I think its locking worked better at least. I think NFSv4's caching rules were also somehow different, but I don't know if they make actimeo=0 unnecessary.
signature.asc
Description: This is a digitally signed message part
