> hardware (or vm?) are you using?
yes, All the crashed OSDs Has the some rocksdb corruption.
Our Cluster has 3 Nodes, and Per Node has
2 * Intel(R) Xeon(R) E5-2620 v3 @ 2.40GHz CPU
4 * 16G DDR4-2133 memory
10 * OSD ( Hitachi 1T 7200 RPM SATA HDD for data, Intel S3500 800G SATA SSD for WAL and DB (WAL Size is 576MB, DB size is 1GB) )
> Also,
> rocksdb:[/clove/vm/clove/ceph/rpmbuild/BUILD/ceph-12.2.0/src/rocksdb/db/compaction_job.cc:1403]
> ....
> it looks like this is a custom build? Are there any changes to the
> source code?
yes, we builded the source code by ourself, but there are not any changes for source code .
Original Mail
Sender: <sage@xxxxxxxxxxxx>;
To: WeiQiaoMiao00105316;
CC: <ceph-users@xxxxxxxxxxxxxx>;
Date: 2017/09/17 01:56
Subject: Re: [ceph-users] osd crash because rocksdb report ‘Compaction error: Corruption: block checksum mismatch’
On Fri, 15 Sep 2017, wei.qiaomiao@xxxxxxxxxx wrote:
>
> Hi，all
>
> My cluster running 12.2.0 with bluestore， we used fio tool with
> librbd ioengine make io test yesterday, and serval osds crash one after
> another.
>
> 3 * node, 30 OSD, 1TB SATA HDD for OSD data, 1GB SATA SSD partition for
> db, 576 MB SATA SSD partition for wal.
>
> ceph options:
>
> bluestore_shard_finishers = true
> mon_osd_prime_pg_temp = false
> mon_allow_pool_delete = true
> mgr_op_latency_sample_interval = 300
All of the crashed OSDs had the same rocksdb corruption error? What kind
of hardware (or vm?) are you using?
Also,
rocksdb:[/clove/vm/clove/ceph/rpmbuild/BUILD/ceph-12.2.0/src/rocksdb/db/compaction_job.cc:1403]
....
it looks like this is a custom build? Are there any changes to the
source code?
Thanks!
sage
>
> Hi，all
>
> My cluster running 12.2.0 with bluestore， we used fio tool with
> librbd ioengine make io test yesterday, and serval osds crash one after
> another.
>
> 3 * node, 30 OSD, 1TB SATA HDD for OSD data, 1GB SATA SSD partition for
> db, 576 MB SATA SSD partition for wal.
>
> ceph options:
>
> bluestore_shard_finishers = true
> mon_osd_prime_pg_temp = false
> mon_allow_pool_delete = true
> mgr_op_latency_sample_interval = 300
All of the crashed OSDs had the same rocksdb corruption error? What kind
of hardware (or vm?) are you using?
Also,
rocksdb:[/clove/vm/clove/ceph/rpmbuild/BUILD/ceph-12.2.0/src/rocksdb/db/compaction_job.cc:1403]
....
it looks like this is a custom build? Are there any changes to the
source code?
Thanks!
sage
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
