Problems running ctl_mboxlist -m on 2.2 back-end

David Mayo D.J.Mayo at bath.ac.uk
Tue Jul 28 08:38:08 EDT 2009


Michael,

Michael Bacon wrote:

> I think I may have seen exactly what you're seeing, and it's a doozy to 
> find, but simple to fix.  If it's what I saw (and it was this EXACT 
> symptom), you just need to rebuild your binaries with the thread-safe 
> switch.  If you're using Sun Studio (we did because of its optimization 
> on the T2 processor), you need to pass "-mt" to the compiler at compile 
> time (CFLAGS), or else Solaris won't set errno properly in a 
> multi-threaded process, and non-blocking I/O will eat itself.  The 
> painful details here:
> 
> http://cyrusimap.web.cmu.edu/archive/message.php?mailbox=archive.info-cyrus&msg=48912 
> 
> The switch on GCC is different (I think it's -mthread), but just make 
> sure you're specifying the thread-safe switch.

This was it exactly - thanks!! The agrument to gcc is in fact -pthreads. 
Better still, I only had to apply this to the MUPDATE server, which 
means I don't need to recompile imapd on the live back-end server.

Allow me to buy you a beer or two if you're ever in the area!

Regards,


Dave.

David Mayo
Networks/Systems Administrator
University of Bath Computing Services, UK

> --On July 23, 2009 1:35:00 PM +0100 David Mayo <D.J.Mayo at bath.ac.uk> wrote:
> 
>>
>> We are looking at upgrading our single 2.2 IMAP server to a Murder setup
>> with a 2.3 back-end server. For the transition we will put the current
>> IMAP server into the Murder and gradually transfer the mailboxes over to
>> the new server using 'xfer'.
>>
>> I have just tested the first stage of the migration by dumping the list
>> of mailboxes on the currently live server and importing that into our
>> test 'currently live' server. The initial ctl_mboxlist -m transfer to
>> the MUPDATE master took ~12 minutes for ~225,000 mail folders (skiplist
>> format).
>>
>> Subsequent attempts to run ctl_mboxlist -m on the server do not work -
>> there is a small flurry of activity at the start according to truss on
>> the local machine and snoop on the MUPDATE server, then nothing happens
>> for exactly 30 minutes and it finally gives up with "couldn't do LIST
>> command on mupdate server".
>>
>> I have restarted the IMAP daemons on both servers and tried converting
>> the mboxlist_db on the back-end from skiplist to berkeley - none of
>> these steps have made any difference. I can run mupdatetest and issue a
>> LIST command which shows plenty of mailboxes.
>>
>> This shouldn't be a problem in itself as long as all the mailbox
>> operations work as expected, however it is a bit of a worry. Both
>> machines are running Solaris 10. The back-end server is running 2.2.12
>> and the front-end server is running 2.2.13. Has anyone experienced this
>> problem and is there a way round it if this command doesn't work?
>>
>> Regards,
>>
>>
>> Dave.
>>
>> David Mayo
>> Networks/Systems Administrator
>> University of Bath Computing Services, UK
>> ----
>> Cyrus Home Page: http://cyrusimap.web.cmu.edu/
>> Cyrus Wiki/FAQ: http://cyrusimap.web.cmu.edu/twiki
>> List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html
> 
> 
> 
> 


More information about the Info-cyrus mailing list