Thanks for all your suggestions. File system optimizations via mount options did not help. Changing to ext3 or creating an ext3 partition is unfortunately not possible. Our current solution is to patch Exim to disable fsync. This reduces the time to thaw to 15 ms which is similar to the performance we get with our first server group on ext3. It's a bit ugly but given the circumstances it seems to work fine. As a precaution we will use the modified Exim binary only for thawing messages. This implies that overall Exim performance may still be slower than it could be.
Regards, Volker -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Jakob Hirsch Sent: Thursday, August 23, 2012 9:53 AM To: exim users Subject: Re: [exim] Exim message thaw performance Volker Schmelich, 22.08.2012 22:15: > Thank you for the tip regarding strace. From a first glance it looks like server 2 loses time between "Writing spool header file" and "LOG: MAIN unfrozen by root". Especially some fstat(), fsync(), close() call lines have higher times displayed when using strace with the -r param. > > E.g. one close() shows 0.000644 on server 1 and the same close() on server 2 shows 0.032874. close() of file read or written to? Maybe you could put the strace outputs (-o /tmp/strace) on some webserver... > So my uneducated guess would be that ext4 is to blame. Is there a way to temporarily make ext4 behave like ext3 and re-run the test? It should be possible to mount the volume as ext3 (unless a non-backwards-compatible ext4 feature was enabled, obviously). Or just create a new ext3 partition, copy your spool there and mount it over /var/spool/exim. There are some ext4 options that may change performance, e.g. noauto_da_alloc, nobarrier, journal_data_writeback. -- ## List details at https://lists.exim.org/mailman/listinfo/exim-users ## Exim details at http://www.exim.org/ ## Please use the Wiki with this list - http://wiki.exim.org/ -- ## List details at https://lists.exim.org/mailman/listinfo/exim-users ## Exim details at http://www.exim.org/ ## Please use the Wiki with this list - http://wiki.exim.org/
