Agreed. I can confirm ceph version 0.60-669-ga2a23cc
(a2a23ccd959f6e7ebe1533b27e7320902624523b) fixes the tcmalloc issue.
root@xxx:~# ceph -m 10.1.0.67 heap stats
mon.btcmalloc heap stats:------------------------------------------------
MALLOC: 25694552 ( 24.5 MiB) Bytes in use by application
MALLOC: + 53272576 ( 50.8 MiB) Bytes in page heap freelist
MALLOC: + 1756920 ( 1.7 MiB) Bytes in central cache freelist
MALLOC: + 5027328 ( 4.8 MiB) Bytes in transfer cache freelist
MALLOC: + 8341936 ( 8.0 MiB) Bytes in thread cache freelists
MALLOC: + 1372312 ( 1.3 MiB) Bytes in malloc metadata
MALLOC: ------------
MALLOC: = 95465624 ( 91.0 MiB) Actual memory used (physical + swap)
MALLOC: + 7954432 ( 7.6 MiB) Bytes released to OS (aka unmapped)
MALLOC: ------------
MALLOC: = 103420056 ( 98.6 MiB) Virtual address space used
MALLOC:
MALLOC: 3760 Spans in use
MALLOC: 34 Thread heaps in use
MALLOC: 8192 Tcmalloc page size
------------------------------------------------
Call ReleaseFreeMemory() to release freelist memory to the OS (via
madvise()).
Bytes released to the
- Mike
On 4/26/2013 4:13 PM, Mark Nelson wrote:
On 04/26/2013 02:43 PM, Sage Weil wrote:
I've merged the double-forward fix. I still have one remaining problem
that I can reproduce with my tests which looks like an overlap of mon
overload forwarded message priorities; I'm testing two possible fixes to
see which are necessary to resolve it.
The memory issues with people using the gitbuilder appear to be because
our gitbuilders weren't linking in tcmalloc. The latest 'next' build
*does* link it now, so if you are one of those people, please re-test!
I don't think either of those problems is related to the sync issues Mike
is seeing, though. I believe that is the final outstanding issue...?
sage
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
fwiw, I'm seeing consistently lower mon usage now. 1-2GB instead of
5-14GB during performance tests.
Mark
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
