Bug #16805
Updated by Kefu Chai over 7 years ago
<pre>
2016-07-24T21:26:58.049 INFO:tasks.thrashosds.thrasher:Reweighting by: {'overage': '1000', 'type': 'reweight-by-utilization', 'max_change': '0.05'}
2016-07-24T21:26:58.050 INFO:teuthology.orchestra.run.mira101:Running: 'sudo adjust-ulimits ceph-coverage /home/ubuntu/cephtest/archive/coverage timeout 120 ceph --cluster ceph osd
reweight-by-utilization 1000 0.05'
</pre>
<pre>
2016-07-24T21:26:58.633 INFO:teuthology.orchestra.run.mira101.stderr:2016-07-25 04:26:58.629902 7f0889763700 1 -- 172.21.8.118:0/3981205993 <== mon.0 172.21.7.112:6789/0 8 ==== mo
n_command_ack([{"prefix": "osd reweight-by-utilization", "oload": 1000, "max_change": 0.05}]=-33 Refusing to reweight: we only have 584164 kb used across all osds! v291) v1 ==== 17
6+0+0 (1169973010 0 0) 0x7f0870001240 con 0x7f0884064ec0
</pre>
see http://pulpito.ceph.com/kchai-2016-07-24_20:41:33-rados-wip-kefu-testing---basic-mira/331659/
the @mon_reweight_min_bytes_per_osd@ by default is 100MB = 102400 kb. 104857600. while we only had 584164 kb ~= 570KB on average. maybe we shoulder lower @mon_reweight_min_bytes_per_osd@ when performing the thrashing test?