Ceph

11/21/2017

09:17 PM bluestore Bug #22061: Bluestore: OSD killed due to high RAM usage

I can confirm this.
Problem informations:
- write only workload
- erasure coding
Really looks like : http:/... red ref

07/20/2017

06:55 PM Messengers Bug #20670: OSD suicide on msgr exceeding fd limit

In the meantime, I found the root cause by using ceph-fuse in interactive mode (not using fstab) and got "Too many op... red ref

07/18/2017

06:33 PM Messengers Bug #20670 (Closed): OSD suicide on msgr exceeding fd limit

On fresh 12.1.0 install (and then 12.1.0-707), with bluestore + fuse + cephfs + ec_overwrites, I got some OSD flapp... red ref

06/22/2017

05:46 PM RADOS Bug #19909: PastIntervals::check_new_interval: assert(lastmap->get_pools().count(pgid.pool()))

Basically yes.
In src/mon/Session.h -> Subscription->next = -1 or 0.
I am learning C++ standard all the way but... red ref

06/21/2017

09:13 AM RADOS Bug #19909: PastIntervals::check_new_interval: assert(lastmap->get_pools().count(pgid.pool()))

Yes, i'm pretty sure it was 12.0.3. But, not on first boot, only after massive failures got me to stale+down PG statu... red ref

06/20/2017

06:45 PM RADOS Bug #19299: Jewel -> Kraken: OSD boot takes 1+ hours, unusually high CPU

Same problem here (fresh 12.0.3). Got OSD's behind by > 5000 maps, it took ~8 hours to get them booted.
Looking in... red ref

06:30 PM RADOS Bug #19909: PastIntervals::check_new_interval: assert(lastmap->get_pools().count(pgid.pool()))

I can confirm the second behavior ("failed to load OSD map for epoch 1") in native installed 12.0.3 (not in productio... red ref

06/16/2017

08:01 PM Messengers Fix #20330 (Resolved): msg : tcp backlog parameter increase

On a quite large cluster, during a cold restart phase, I noticed from tcp metrics some "ListenOverflows" and "Liste... red ref