imap: operation timed out

[Sneeper]

With Scalix 10 (I don't remember seeing this on previous versions but I wasn't using the previous version for that long), I'm getting messages from my IMAP client (thunderbird) that an operation timed out whenever I do something big -- like move 1000 messages from one folder to another.

omshowlog shows these messages:

Code: Select all

WARNING                        Internet Mail (Incoming      ) 03.17.06 06:57:07
[OM.UX 1401] Cannot read message data.
        <- tf_AppendRecord
        -> GetMailText
        <- GetMailText
        -> GetMailText
        <- GetMailText
        -> ux_DecrementNesting
        <- ux_DecrementNesting
        <- ux_InParseMimeMultiPart
        -> im_UnlinkItem
        -> im_ItemRef2FName
        <- im_ItemRef2FName
        <- im_UnlinkItem
        <- ux_InParseMimeContent
        <- /build/10.0.0.175/src/bin/ux/ux_in.c:4672[102,1401]
        <- /build/10.0.0.175/src/bin/ux/ux_in.c:4767[102,1401]
        <- /build/10.0.0.175/src/bin/ux/ux_in.c:2187[102,1401]


But I'm not sure if they are related or something totally different. I'm more concerned about fixing the operation time outs than these error messages if they aren't related.

Any ideas?

[Sneeper]

To add to this, I'm actually getting 'connection timed out' even on just logging in. It's not just on large operations. It's not every time, but it's often. :(

[florian]

After the upgrade, did you run

Code: Select all

omtidyallu -M

This will regenerate the cached copy of the MIME version of your messages; this will greatly improve IMAP performance. The MIME rendering format has changed with Scalix 10 - if you don't run this command (see release notes or install guide), your MIME cache will be updated on access which is slow and can lead to timeouts.

I'm using Thunderbird against my 20000 item Inbox and never get any timeouts.

Florian.

[Sneeper]

No, I didn't, but I'll try that. Thanks for the suggestion. I'll see if I get timeouts after that.

[Sneeper]

I'm sorry to say that this did not help.. I still get time outs for large operations. :(

[jch]

That error message is pretty well independent of a timeout being reported by Thunderbird.

Recent versions of Thunderbird have taken to getting bored with an operation that they think is taking too long and will claim a timeout error. We've recently had problems like this on one of our development servers (pretty well fixed now, you'll be glad to hear) but there are various things that excerbate the problem.

When you first log in after the server has started or if you log in again after you've been logged out for quite some time, a server process starts up to cache information to make subsequent operations much quicker. This isn't any more work that used to be done (one way or another) but it can take a while on a heavily loaded system.

If you can log into the server with ssh or similar and run "top" when you attempt a login that you suspect is going to fail, are you seeing what looks like an abnormally high load for your server? Has the server (or at least the scalix server) recently started?

Once you've logged in and you're getting timeout problems with large operations, can you do the same sort of thing? Does top show a lot of I/O wait? (Some versions show a percentage under "iowait" some just say "wa"). A very high load? The type of disks that you have will make a big difference -- if you got one IDE disk on a built-in controller(*) then you won't get the I/O throughput that you would with an array of mirrored+striped 15K SCSI disks (but it'll be a lot cheaper). It basically comes down to system load.

jch


(*) Some Dell servers have a setting in the BIOS to limit IDE disks to a 33Mhz bus -- if you put an 80 conductor cable in you can up the speed to 66Mhz or even 100Mhz and, needless to say, it makes a big difference.