-
Notifications
You must be signed in to change notification settings - Fork 5.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
mempool: expose perfcounters for mempools; other perf counter cleanup #14982
Conversation
Lots of this: 2017-05-14T18:16:27.589 INFO:tasks.ceph.osd.2.smithi176.stderr:/build/ceph-12.0.1-2390-g52f2336/src/common/mempool.cc: In function 'size_t mempool::pool_t::allocated_bytes() const' thread 7fbf9a7a2700 time 2017-05-14 18:16:27.594220 2017-05-14T18:16:27.589 INFO:tasks.ceph.osd.2.smithi176.stderr:/build/ceph-12.0.1-2390-g52f2336/src/common/mempool.cc: 66: FAILED assert(result >= 0) 2017-05-14T18:16:27.590 INFO:tasks.ceph.osd.2.smithi176.stderr: ceph version 12.0.1-2390-g52f2336 (52f2336af234f3d5e399ed60a3630166dcd593a0) 2017-05-14T18:16:27.590 INFO:tasks.ceph.osd.2.smithi176.stderr: 1: (ceph::__ceph_assert_fail(char const*, char const*, int, char const*)+0x10e) [0x7fbf9fe930ce] 2017-05-14T18:16:27.590 INFO:tasks.ceph.osd.2.smithi176.stderr: 2: (()+0x9bc1d5) [0x7fbf9fe871d5] 2017-05-14T18:16:27.590 INFO:tasks.ceph.osd.2.smithi176.stderr: 3: (CephContext::_refresh_perf_values()+0x78) [0x7fbf9ff68658] 2017-05-14T18:16:27.591 INFO:tasks.ceph.osd.2.smithi176.stderr: 4: (CephContextServiceThread::entry()+0x160) [0x7fbf9ff6d620] 2017-05-14T18:16:27.591 INFO:tasks.ceph.osd.2.smithi176.stderr: 5: (()+0x8184) [0x7fbf9d9ca184] 2017-05-14T18:16:27.592 INFO:tasks.ceph.osd.2.smithi176.stderr: 6: (clone()+0x6d) [0x7fbf9caba37d] we may need to drop the assert as it's racy? Or we tested a buggy version of the patch. |
232d790
to
e6ea4b8
Compare
Found this PR here: https://trello.com/c/xoMYLms8/12-expose-mempool-metrics |
This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. |
rgw and rbd_mirror were enabling a perfcounter monitoring the heartbeatmap. Enable this unconditionally for all ccts. This is simpler and more useful! Add a CINIT_FLAG to avoid starting them up. This is only used by the unit tests which do not expect all the cct crap and we don't want to break every time we add a new cct counter. That means removing the CephContextPerfCounters test. Signed-off-by: Sage Weil <sage@redhat.com>
Report bytes and items for each mempool. Signed-off-by: Sage Weil <sage@redhat.com>
e6ea4b8
to
c3c1393
Compare
rebased |
If we are adding up the shards, and race with some allocations and deallocations, we might get a negative sum. E.g., with 2 shards, intially [0,0], - A: read shard 0 (0) - B: allocate object on shard 0 (now 1) - C: dallocate object on shard 1 (now -1) - A: read shard 1 (-1) - A: sum is -1 Signed-off-by: Sage Weil <sage@redhat.com>
* refs/pull/14982/head: common/mempool: tolerate sum < 0 common/ceph_context: add perf counters for all mempools common/ceph_common: add heartbeat perf counters unconditionally Reviewed-by: Josh Durgin <jdurgin@redhat.com>
No description provided.