1 Topic by ketan.aagja

  • ketan.aagja
  • Member
  • Offline
  • From: UK
  • Registered: 2012-05-15
  • Posts: 252

Topic: Unexpected error in postfix/dovecot after running fine for 3 days

==== Provide required information ====
- iRedMail version and backend (LDAP/MySQL/PGSQL): iRedAdmin-Pro-LDAP-1.7.2
- Linux/BSD distribution name and version: Centos 5.8
- Any related log? Log is helpful for troubleshooting.
====

Hi,

After running flawlessly for 3 days today one of my busiest E-mail id stopped working, thunderbird application now asking "sending of password did not succeed. Mail server responded:+" upon checking dovecot log I found out:

May 23 20:26:31 POP3(sonicwall@mydomain.com): Error: fcntl(write-lock) locking failed for file /mnt/glusterfs/vmail1/mydomain.com/x/x/x/sonicwall-2012.05.20.10.56.11//Maildir/dovecot.index.log: Transport endpoint is not connected
May 23 20:26:31 POP3(sonicwall@mydomain.com): Error: mail_index_wait_lock_fd() failed with file /mnt/glusterfs/vmail1/mydomain.com/x/x/x/sonicwall-2012.05.20.10.56.11//Maildir/dovecot.index.log: Transport endpoint is not connected
May 23 20:26:31 POP3(sonicwall@mydomain.com): Warning: Our dotlock file /mnt/glusterfs/vmail1/mydomain.com/x/x/x/sonicwall-2012.05.20.10.56.11//Maildir/dovecot-uidlist.lock was overridden (locked 3611 secs ago, touched 3611 secs ago)
May 23 20:26:31 POP3(sonicwall@mydomain.com): Error: Couldn't init INBOX: Internal error occurred. Refer to server log for more information. [2012-05-23 19:26:20]

Is there any solution?

----

Spider Email Archiver: On-Premises, lightweight email archiving software developed by iRedMail team. Supports Amazon S3 compatible storage and custom branding.

2 Reply by ZhangHuangbin

  • ZhangHuangbin
  • ZhangHuangbin
  • iRedMail Developers
  • Offline
  • Registered: 2009-05-06
  • Posts: 31,082

Re: Unexpected error in postfix/dovecot after running fine for 3 days

No idea at all. Found several same reports in Dovecot mailing list (with Google), but seems it was fixed in Dovecot-2.0.1 and later versions. On CentOS 5, it should be Dovecot-1.2.x, not sure whether it was fixed or not.

Another possible solution: Store mailbox index files on local file system (ext3/ext4/reiserfs/etc) instead of GlusterFS. To achieve this, update "mail_location" setting in dovecot.conf:

# ---- OLD SETTING
# Maildir format and location.
#mail_location = maildir:/%Lh/Maildir/:INDEX=/%Lh/Maildir/

# ---- NEW SETTING
mail_location = maildir:/%Lh/Maildir/:INDEX=/some/place/%Ld/%Ln/

Restarting Dovecot service is required.

3 Reply by ketan.aagja

  • ketan.aagja
  • Member
  • Offline
  • From: UK
  • Registered: 2012-05-15
  • Posts: 252

Re: Unexpected error in postfix/dovecot after running fine for 3 days

hi Zhang,

Upon Google I found out that in dovecot.conf adding lock_method = dotlock resolves the issue and I tried and got it resolved. Yet from last night email application has not asked me for the password and have not observed error logs in dovecot.log also.

I just want to know does it affect to glusterfs working?

4 Reply by ZhangHuangbin

  • ZhangHuangbin
  • ZhangHuangbin
  • iRedMail Developers
  • Offline
  • Registered: 2009-05-06
  • Posts: 31,082

Re: Unexpected error in postfix/dovecot after running fine for 3 days

ketan.aagja wrote:

I just want to know does it affect to glusterfs working?

Not sure. I don't have a GlusterFS, so you have to try it yourself.