Ceph : Issueshttps://tracker.ceph.com/https://tracker.ceph.com/favicon.ico2018-04-14T05:50:45ZCeph
Redmine RADOS - Support #23719 (New): Three nodes shutdown,only boot two of the nodes,many pgs down.(host...https://tracker.ceph.com/issues/237192018-04-14T05:50:45Zjunwei liao
<p>The interval mechanism of PG will cause a problem in the process of cluster restart.If I have 3 nodes(host failure-domain/ec 2+1) in my cluster,after I close the 3 nodes, I only boot 2 of them,many pgs down.Theoretically,host failure-domain/ec 2+1, 3 nodes have 2 online to guarantee PG active.The reason is because of the PG interval mechanism.</p>
<p>If pg1.1 have acting set [1,2,3], I power down osd.3 first,then power down osd.2. In case I boot osd.3 first,boot osd.2 second,then pg can be active;in case I boot osd.2 first,boot osd.3 second,then pg can be down. When I close the 3 nodes in the cluster at the same time, all osd power down disorder.If I only boot 2 of them,then cause a lot of pgs down.The reliability of the cluster will be greatly affected.</p> RADOS - Bug #23235 (New): The randomness of the hash function causes the object to be inhomogeneo...https://tracker.ceph.com/issues/232352018-03-06T03:03:46Zjunwei liao
<p>The randomness of the ceph_str_hash_rjenkins hash function causes the object to be inhomogeneous to the PG.The result is that each OSD utilization ratio is uneven.The highest utilization ratio of OSD is 10% higher than the lowest utilization ratio of OSD that causes the utilization rate of the whole cluster capacity is not high.</p>
<p>In addition, I use the pg-upmap command to make the same number of PG on each OSD.However, when a cluster has a OSD utilization ratio of 95%,the highest utilization ratio of OSD is still about 10% higher than the lowest utilization ratio of OSD.It can be explained that the inhomogeneity of the OSD utilization is not the crush algorithm but the inhomogeneity of the obejct to PG.</p> RADOS - Bug #23233 (Duplicate): The randomness of the hash function causes the object to be inhom...https://tracker.ceph.com/issues/232332018-03-06T02:34:07Zjunwei liaoCeph - Bug #22757 (Closed): 色方式https://tracker.ceph.com/issues/227572018-01-23T02:31:41Zjunwei liaobluestore - Bug #21531 (Can't reproduce): BlueStore::TwoQCache::_trim:BlueStore cache can not be ...https://tracker.ceph.com/issues/215312017-09-25T03:03:54Zjunwei liao
<p>My cluster configuration is 10 BlueStore OSDs, ec overwrite 8+2, BlueStore_cache_size(only onode cache) = 16MB (for reduce memory consumption).I created 30 RBD image and wrote these by librbd of python.After a while,I checked dump_mempools,found the sum of bluestore_cache_onode and bluestore_cache_other are more than 200MB.So I opened debug_bluestore to 20,the log is like as:<br />2017-09-21 21:12:10.588694 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:c8af972f:::*inc_osdmap*.80:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588700 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:374ce826:::*osdmap*.80:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588703 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:c551972f:::inc_osdmap.81:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588706 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:3dcce826:::osdmap.81:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588708 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:ce51972f:::inc_osdmap.82:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588710 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:32cce826:::osdmap.82:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588714 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:c0d1972f:::inc_osdmap.83:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588716 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:380ce826:::osdmap.83:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588718 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:cb11972f:::inc_osdmap.84:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588721 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:358ce826:::osdmap.84:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588724 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:c611972f:::inc_osdmap.85:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588726 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:3e8ce826:::osdmap.85:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588728 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:cc91972f:::inc_osdmap.86:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588730 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:306ce826:::osdmap.86:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588732 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:c371972f:::inc_osdmap.87:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588735 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:3bece826:::osdmap.87:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588738 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:c9f1972f:::inc_osdmap.88:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588740 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:36ece826:::osdmap.88:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588742 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:c4f1972f:::inc_osdmap.89:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588744 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:3c2ce826:::osdmap.89:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588747 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:c1b1972f:::inc_osdmap.90:0# has 2 refs; skipping<br />2017-09-21 21:12:10.588750 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:39536826:::osdmap.90:0# has 2 refs; skipping<br />······<br />2017-09-21 21:12:10.588758 7fe3a5bf2700 20 bluestore.2QCache(0x7fe3c185c2a0) _trim #-1:39536826:::superblock# has 2 refs; skipping</p>
<p>Makebe all Onode of coll_t::meta will not be unref? but why?</p>