Bug #46411
Updated by Kefu Chai almost 4 years ago
* 1.anomalies When the ceph cluster is idle,you can found that the disks associated to osds have small write io every bluefs_balance_interval seconds. just like below: <pre><code class="c"> Device: rrqm/s wrqm/s r/s w/s rMB/s wMB/s avgrq-sz avgqu-sz await r_await w_await svctm %util sdb 0.00 0.00 0.00 4.00 0.00 0.01 4.00 0.00 0.00 0.00 0.00 0.00 0.00 sdc 0.00 0.00 0.00 4.00 0.00 0.01 4.00 0.00 0.00 0.00 0.00 0.00 0.00 sdf 0.00 0.00 0.00 4.00 0.00 0.01 4.00 0.00 0.00 0.00 0.00 0.00 0.00 sdk 0.00 0.00 0.00 4.00 0.00 0.01 4.00 0.00 0.00 0.00 0.00 0.00 0.00 </code></pre> * 2.reason This abnormal phenomenon is caused by the pr(https://github.com/ceph/ceph/pull/26866 ),so pr(https://github.com/ceph/ceph/pull/26866),so ceph version of v12.2.12~~v12.2.13, v13.2.5 ~~ v13.2.9 has the problem. The reason is that it will lead to an empty RocksDB transaction submite to RocksDB even on an idle system,the empty transaction lead to the small write. More details can be seen in the discussion in that pr. * 3.solution check if the transaction is empty,then it's will not be submitted to the rocksdb.