imapsync very poor performance

Christian Hügel christian.huegel at stonebyte.de
Tue Dec 3 14:16:30 CET 2013


Hi Matt,

just finished sync-ing. 8GB in 5 hours....thats ok for me.


Am 03-12-2013 14:00, schrieb Matt Moldvan:
> In case last access time is useful information, you probably only need
> noatime on wherever the mailboxes are stored (/var?).  Our Cyrus 
> databases
> at work went from 2 hours to rebuild to about 2 minutes using that 
> option.
> I'm curious, what was the speedup you experienced?
> On Dec 3, 2013 6:09 AM, "Christian Hügel" 
> <christian.huegel at stonebyte.de>
> wrote:
> 
>> Hi Matt, hi Franz,
>> 
>> I have remounted all partitions with the option *noatime*
>> and now it works!!!
>> thank you so much for your support.
>> 
>> Cheers,
>> 
>> Christian
>> 
>> Am 03-12-2013 03:09, schrieb Matt Moldvan:
>> 
>>> In your mount options, try using the "noatime" option.  You can 
>>> probably
>>> apply this with a "mount -o remount,noatime /dev/something
>>> /mnt/somewhere",
>>> then restart dovecot for fun and try again.  Is the original system 
>>> you
>>> are
>>> having trouble with also a VM?  If so, change to scheduler for the 
>>> disk
>>> drive it's mounted on to "noop":
>>> 
>>> [someone at somewhere #] cat /sys/block/somedrive/queue/scheduler
>>> noop deadline [cfq]
>>> [someone at somewhere #] echo noop > 
>>> /sys/block/somedrive/queue/scheduler
>>> [someone at somewhere #] cat /sys/block/somedrive/queue/scheduler
>>> [noop] deadline cfq
>>> 
>>> 
>>> On Mon, Dec 2, 2013 at 7:26 PM, Franz Skale <i.bin at dah.am> wrote:
>>> 
>>>  Check your filesystem and I/O errors in dmesg. Check for the 
>>> maildirs.
>>>> A punch of posibilities.
>>>> 
>>>> 
>>>> Rgds Franz
>>>> 
>>>> Telepathisch gesendet
>>>> 
>>>> Am 03.12.2013 um 00:24 schrieb Christian Hügel <
>>>> christian.huegel at stonebyte.de>:
>>>> 
>>>> > Thx, I get a bunch of errors like this one:
>>>> >
>>>> > [pid  6801] <... read resumed> "121 SELECT INBOX.User\r\n", 1639) = 24
>>>> > [pid  6800] <... read resumed> 0x17189520, 5) = -1 EAGAIN (Resource
>>>> > temporarily unavailable)
>>>> > [pid  6801] close(14 <unfinished ...>
>>>> > [pid  6800] epoll_wait(8,  <unfinished ...>
>>>> > [pid  6801] <... close resumed> )       = 0
>>>> >
>>>> >
>>>> > Am 02.12.2013 23:46, schrieb Franz Skale:
>>>> >> strace -f -s 16384 dovecot -F 2>/tmp/dovecot.debug
>>>> >
>>>> > --
>>>> 
>>> 
>> 


More information about the users mailing list