General

Profile

Marcin W

  • Registered on: 10/26/2019
  • Last connection: 11/06/2019

Issues

Activity

03/26/2020

08:53 AM bluestore Bug #38745: spillover that doesn't make sense
Hi Igor,
So, your recommendation is to create a volume which can serve enough space for levels 1-3, compaction and...

03/24/2020

11:56 AM bluestore Bug #38745: spillover that doesn't make sense
BTW. These are the default RocksDB params IIRC (in GB):
base=256, multiplier=10, levels=5
0= 0.25
1= 0,25
2= 2,...
11:47 AM bluestore Bug #38745: spillover that doesn't make sense
@Igor,
I've read somewhere in docs that recommended DB size is no less that 5% of block. IMO if we consider either...

12/13/2019

02:07 PM bluestore Bug #42223: ceph-14.2.4/src/os/bluestore/fastbmap_allocator_impl.h: 750: FAILED ceph_assert(avail...
Due to this bug, a had to mitigate it with setting RocksDB param min_write_buffer_number_to_merge to 2. Since 14.2.5 ...

11/13/2019

08:27 PM bluestore Bug #42223: ceph-14.2.4/src/os/bluestore/fastbmap_allocator_impl.h: 750: FAILED ceph_assert(avail...
@Igor,
Well done!
Would it also explain problem I had with 003808.sst(30MB) where all others were 16MB as they w...

11/10/2019

12:02 PM bluestore Bug #42223: ceph-14.2.4/src/os/bluestore/fastbmap_allocator_impl.h: 750: FAILED ceph_assert(avail...
Dev guys,
I haven't checked the source code for handling KV store but is it possible that some parts use buffered ...

11/02/2019

04:57 PM bluestore Bug #42223: ceph-14.2.4/src/os/bluestore/fastbmap_allocator_impl.h: 750: FAILED ceph_assert(avail...
I ran bluefs-export export twice on the same OSD and data MD5 hashes are identical for both copies.
04:31 PM bluestore Bug #42223: ceph-14.2.4/src/os/bluestore/fastbmap_allocator_impl.h: 750: FAILED ceph_assert(avail...
Same here. Although, I have another OSD with same issues.

10/30/2019

05:31 PM ceph-volume Bug #36307: allow arbitrary OSD IDs to be used when creating+preparing
...
02:41 PM bluestore Bug #42223: ceph-14.2.4/src/os/bluestore/fastbmap_allocator_impl.h: 750: FAILED ceph_assert(avail...
Tobias,
Perhaps the below helps to avoid corruption:
I also had https://tracker.ceph.com/issues/20381 in logs:
...

Also available in: Atom