Hello,
I have a dedicated server: Intel(R) Xeon(R) CPU X3470 @ 2.93GHz Quad Core
8GB RAM and 2 x 1 TB SATA Hard Drives
Centos 5.8 x32 bits
I've noticed an issue with Horde Webmail performance when an user access a very large mailbox (>1gb) or with a lot of messages (>1000).
The load average starts to grow because of a high CPU usage of an IMAPD process, that sounds normal because in fact imap-courier starts to read the mailbox but the load average can go up to 12.00 or 14.00 causing the server to become sluggish and non responsive.
The process can get stuck in some kind of cycle and that brings sometimes Qmail or Apache down (when multiple large mailbox users access Horde at the same time).
Doing a little research with lsof | grep pID (where pID is the imapd process) I can get the user mailbox, it always shows a 127.0.0.1 connection, so that's a normal Horde Webmail access.
The load average can get even worse, if the user does some message sorting by date, I guess imap-courier has to re-read all the mailbox again, and if it's way to heavy or with tons of messages, consumes a lot of resources.
Anyone with this kind of issue?
Is there any way to tweak horde/imap-courier to assign maybe more RAM or the opposite, limit those processes?
Maybe going to x64 bits can handle better the imap processes?
Maybe use Atmail instead of Horde? I haven't tested if Atmail does the same.
Or maybe going through another completely different webmail service such as Roundcube?
Greetings to everyone and I hope Parallels support can't take a look to this issue.
Fernando I.
I have a dedicated server: Intel(R) Xeon(R) CPU X3470 @ 2.93GHz Quad Core
8GB RAM and 2 x 1 TB SATA Hard Drives
Centos 5.8 x32 bits
I've noticed an issue with Horde Webmail performance when an user access a very large mailbox (>1gb) or with a lot of messages (>1000).
The load average starts to grow because of a high CPU usage of an IMAPD process, that sounds normal because in fact imap-courier starts to read the mailbox but the load average can go up to 12.00 or 14.00 causing the server to become sluggish and non responsive.
The process can get stuck in some kind of cycle and that brings sometimes Qmail or Apache down (when multiple large mailbox users access Horde at the same time).
Doing a little research with lsof | grep pID (where pID is the imapd process) I can get the user mailbox, it always shows a 127.0.0.1 connection, so that's a normal Horde Webmail access.
The load average can get even worse, if the user does some message sorting by date, I guess imap-courier has to re-read all the mailbox again, and if it's way to heavy or with tons of messages, consumes a lot of resources.
Anyone with this kind of issue?
Is there any way to tweak horde/imap-courier to assign maybe more RAM or the opposite, limit those processes?
Maybe going to x64 bits can handle better the imap processes?
Maybe use Atmail instead of Horde? I haven't tested if Atmail does the same.
Or maybe going through another completely different webmail service such as Roundcube?
Greetings to everyone and I hope Parallels support can't take a look to this issue.
Fernando I.