|
From: Jesse C. S. <jsm...@ga...> - 2006-10-07 04:26:26
|
Currently I'm using Sendmail with the mbox format. What you say does make sense though. Towards the end of the year I did notice our email traffic increasing so it would be only reasonable that the old way of doing things could cause us issues with incoming mail on the rise. Doing a quick search on google I noticed that doing Maildir shouldn't be a huge problem with some minor mods to promailrc. However what I don't understand is does that mean new mail goes into /var/spool/mail/username/<messages here> or would it be ~username/mail/inbox/<messages here>? Lastly, as far as the conversion is usermin cabable of reading two different formats at a time? I ask because I wonder can I just convert say all the inboxs first, wait to see what it does, and then convert all of their personal folders. Or do I need to do both sides of the puzzle at the same time? We currently are using Dovecot for the Pop3 / Imap side of things and it appears to be really easy to convert to the different setting. Thanks Jaime, bTw. -Jesse Jamie Cameron wrote: > On 6/Oct/2006 12:06 Jesse C. Smillie wrote .. > >> Hello once again everyone. Since the new school year started we have >> been noticing increasingly high I/O usage on our servers. When I run >> the top command quite often you will see the WA status swing from really >> low up into the 70-80 even 90% range for up to a min and then drop >> again. I spend a little time trying to track down the problem and have >> noticed the following: >> >> When I/O is really high I always see the following line in "ps aux | >> grep D": >> 1229 14619 0.8 0.2 37272 16520 ? D 14:43 0:00 >> /usr/local/usermin-1.230/mailbox/delete_mail.cgi >> >> (Note: User ID is normally not the same one when polled) >> >> Sometimes I also see the index.cgi file too, but no where near as much >> as the above. >> >> A coworker in our department had made mention that maybe whats happening >> has something do with the process that usermin uses to remove mail from >> a mailbox when its deleted. Something about compacting the box during >> the delete instead of doing it like Thunderbird and waiting to compact >> until space can be saved. I guess case being all the reading and >> rewriting required to removed the dead message from the box. >> >> I don't recall having this problem that bad last year. On occassion >> when almost everyone was logged in things would get a little screwy, but >> not like this. >> >> General information about our config: >> Server: >> Processors: Dual Opteron 242's >> Ram: 6Gig ECC DDR >> HDs: 1.2 TB 3ware Raid Array >> OS: Slamd 64 10.2b (Slackware more or less) >> Kernel: 2.6.18 >> Usermin Version: 1.230 (This was happening in the last version too >> however. I just upgraded two days ago to see if the problem went away) >> >> Unfortunately dealing with I/O problems isn't my speciality so I could >> be barking up the wrong tree here. Any suggestions would be greatly >> appriciated as always... >> > > What format are your mailboxes in? If they are in mbox format (one file per > user), when Usermin deletes email it will re-write out the entire mail > file with the removed emails taken out. As you might expect, this could cause > substantial disk IO on large mail files.. > > The solution I used was to switch to Maildir format, in which each email is stored > in a separate file. Deleting from this format is much quicker.. > > - Jamie > > ------------------------------------------------------------------------- > Take Surveys. Earn Cash. Influence the Future of IT > Join SourceForge.net's Techsay panel and you'll get the chance to share your > opinions on IT & business topics through brief surveys -- and earn cash > http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV > - > Forwarded by the Webmin mailing list at web...@li... > To remove yourself from this list, go to > http://lists.sourceforge.net/lists/listinfo/webadmin-list > > ------------------------------------------------------------- > This mail was scanned by a trial version of BitDefender > For more informations please visit http://www.bitdefender.com > > > ------------------------------------------------------------- > |