High cpu usage for imapd processes on 2.4.12 (Invoca rpm)

Simon Matter simon.matter at invoca.ch
Fri Nov 11 11:51:14 EST 2011


> Hi,
>
> I've got an imap server being hammered by a program that is constantly
> connecting via IMAP, searching, reading and deleting emails. At a very
> rough estimate there'll be 10 to 20 new emails every minute that
> arrive, are checked and deleted.
>
> The original machine is running 2.3.16 without replication and the
> average load rarely went over 1. On the new server running 2.4.12
> (details below) I'm seeing the average load hover around 5 constantly.
> The new box is running replication and this appears to be set up
> correctly as the logs look pretty clean. top shows CPU percentage for
> each imapd process running quite high. The largest mailbox contains
> about 3000 emails and is usually at 50 to 60% CPU.
>
> I/O for the machine looks fine, I'm not seeing any iowait or high
> usage with iostat. It's using local disks so not something I would
> expect to be a problem.
>
> Can anyone suggest where I could start looking to find the cause? Or
> point me in the direction of configuration options that are known to
> have such an effect after upgrading between the two version?
>
> Thanks
> Guy
>
>
> New server:
> name       : Cyrus IMAPD
> version    : v2.4.12-Invoca-RPM-2.4.12-2el5_2 63b65d1c 2011-10-04
> vendor     : Project Cyrus
> support-url: http://www.cyrusimap.org
> os         : Linux
> os-version : 2.6.18-238.el5
> environment: Built w/Cyrus SASL 2.1.22
>              Running w/Cyrus SASL 2.1.22
>              Built w/OpenSSL 0.9.8e-fips-rhel5 01 Jul 2008
>              Running w/OpenSSL 0.9.8e-fips-rhel5 01 Jul 2008
>              Built w/zlib 1.2.3
>              Running w/zlib 1.2.3
>              CMU Sieve 2.4
>              TCP Wrappers
>              mmap = shared
>              lock = fcntl
>              nonblock = fcntl
>              idle = idled
>
> imapd.conf:
> admins: cyradmin cyradmin-repl
> configdirectory: /var/lib/imap
> delete_mode: immediate
> duplicate_db: skiplist
> expunge_mode: default
> hashimapspool: 1
> improved_mboxlist_sort: 1
> lmtp_downcase_rcpt: 1
> munge8bit: 0
> partition-default: /var/spool/imap
> ptscache_db: skiplist
> sasl_pwcheck_method: saslauthd
> sendmail: /usr/sbin/sendmail
> sievedir: /var/lib/imap/sieve
> statuscache_db: skiplist
> sync_authname: cyradmin-repl
> server101_sync_host: 10.10.10.10
> server201_sync_host: 10.11.10.11
> sync_log: 1
> sync_log_channels: server101 server201
> sync_password: replpass
> tls_ca_file: /etc/ssl/certs/domain.com.ca-bundle
> tlscache_db: skiplist
> tls_cert_file: /etc/ssl/certs/domain.com.crt
> tls_key_file: /etc/ssl/certs/domain.com.key
> unix_group_enable: 0
> normalizeuid: 1

Hi,

I've not seen any such issues on my 2.4.x systems after upgrading from
2.3.x and I'm also wondering what's going on.
So, is it the same hardware, os and everything and only cyrus changed? And
then of course I don't know about replication how it affects it.

What comes to mind is, did you try with "expunge_mode: delayed"? That can
make quite a difference IIRC.

Regards,
Simon



More information about the Info-cyrus mailing list