mon/PGMonitor: max avail is 0 if some OSD's weight is 0 #6660
Conversation
|
could you pls change the title to mon/PGMonitor: max avail is 0 if some OSD's weight is 0 Rest looks good |
chengyli
changed the title
|
Thanks Xiaoxi. |
|
LTGM @tchaikov , is that OK for a QA run? as Cinder is using "ceph df" to get available space of Ceph, this bug will take down cinder. |
|
Anyone knows the reason of this Jenkin build failure? Looks that this is not related with the code change. First time build. Skipping changelog. |
|
ping @dachary |
|
@chengyli it timedout after one hour (usually takes max 30 minutes). There unfortunately is not much information in this case and you should try run-make-check.sh locally to reproduce the problem. Feel free to rebase and repush to trigger the bot one more time, to verify this is not a false negative. |
In get_rule_avail(), even p->second is 0, it's possible to be used as divisor and quotient is infinity, then is converted to an integer which is negative value. So we should check p->second value before calculation. It fixes BUG ceph#13840. Signed-off-by: Chengyuan Li <chengyli@ebay.com>
|
@dachary thanks for the instructions. if test -d "ceph-9.2.0"; then find "ceph-9.2.0" -type d ! -perm -200 -exec chmod u+w {} ';' && rm -rf "ceph-9.2.0" || { sleep 5 && rm -rf "ceph-9.2.0"; }; else :; fi |
|
lgtm with a rados qa run |
mon: fix ceph df pool available calculation for 0-weighted OSDs Reviewed-by: Kefu Chai <kchai@redhat.com>
In get_rule_avail(), even p->second is 0, it's possible to be used
as divisor and quotient is infinity, then is converted to an integer
which is negative value.
So we should check p->second value before calculation.
Signed-off-by: Chengyuan Li chengyli@ebay.com