| Summary | Performance problems after upgrade to 4.1.4 |
| Queue | IMP |
| Queue Version | 4.1.4 |
| Type | Bug |
| State | Not A Bug |
| Priority | 2. Medium |
| Owners | |
| Requester | robin.west (at) dal (dot) ca |
| Created | 08/15/2007 (6611 days ago) |
| Due | |
| Updated | 09/05/2007 (6590 days ago) |
| Assigned | 08/16/2007 (6610 days ago) |
| Resolved | 09/05/2007 (6590 days ago) |
| Github Issue Link | |
| Github Pull Request | |
| Milestone | |
| Patch | No |
Priority ⇒ 2. Medium
it's under load is who's going to be able to do the best diagnosing.
You might be better off on the mailing lists with this one.
same server? I ask because the problem could be datatree/share
performance issues and not IMAP if the server with the high load is
also the Horde or DB server.
dedicated servers. Horde is running on six app servers which use an
imap proxy to connect to the IMAP server. Only the IMAP server was
overloaded, the others showed no significant change in performance
after the upgrade.
I'm wondering if it's either (a) the virtual inbox making heavier use
of IMAP or (b) changes to the way Ingo accesses IMAP when doing its
filtering work. But these are just guesses really. Let me know if
there's any other information I can provide that might give a clue.
Thanks again.
-Robin
State ⇒ Feedback
server? I ask because the problem could be datatree/share performance
issues and not IMAP if the server with the high load is also the Horde
or DB server.
I can't make anything out of those IMAP logs, fwiw.
New Attachment: new_webmail.log
State ⇒ Unconfirmed
New Attachment: old_webmail.log
Queue ⇒ IMP
Summary ⇒ Performance problems after upgrade to 4.1.4
Type ⇒ Bug
Priority ⇒ 3. High
severe performance problems on our IMAP server. Backing out the
upgrade resolved the issue and subsequent tests confirmed that the new
Webmail is the source of the load, although we're not sure what it is
doing. I'm hoping you can give us some input or suggestions as to
where to look to resolve the problem.
Our IMAP server is Cyrus IMAP4 2.1.11 running on Red Hat AS3. Normally
it handles around 2000 concurrent sessions and the load average is
between 7 and 20. After the upgrade, concurrent sessions were as
normal but the load went up to over 700.
We are running the following on six application servers (all Red Hat
AS3, PHP 4.3.2):
Horde 3.1.4 (upgraded from 3.0.4)
IMP 4.1.4 (upgraded from 4.0.3)
Ingo 1.1.3 (upgraded from 1.0.1)
Turba 2.1.4 (upgraded from 2.0.2)
Backend db is Postgres 7.3.19-RH and we also use LDAP as an address
book source.
I am attaching two trace files our sysadmins took from our IMAP
server, showing the IMAP commands generated by the same sequence of
actions in both the old installation and the new one. It's not very
readable so I don't know if it's much help.
We've also had 2 or 3 reports of individuals whose inboxes were empty
when they logged in after the upgrade. I'm still investigating those
so I don't have any info yet...
Thanks in advance for any suggestions you can offer!
-Robin