Hi everyone,
I'm looking for some assistance diagnosing a recurring crash affecting a single OSD.
assert_file: ./src/os/bluestore/BlueStore.cc
assert_func: void BlueStore::_txc_apply_kv(TransContext*, bool)
assert_thread_name: bstore_kv_sync
stack_sig:
4bf8806194e49b144d67a8ec58abc1e8997fbbe795691d6d11b5dae2a4faa140
Backtrace:
BlueStore::_txc_apply_kv()
BlueStore::_kv_sync_thread()
BlueStore::KVSyncThread::entry()
The OSD aborts after the failed assertion.
PutCF(...)
PutCF(...)
PutCF(...)
followed immediately by
FAILED ceph_assert(r == 0)
There are no preceding messages such as:
Thanks.
I'm looking for some assistance diagnosing a recurring crash affecting a single OSD.
Environment
- Proxmox VE 9.2.5
- Ceph Squid
- Current version: 19.2.4
- Previous version where the same crash occurred: 19.2.3
- Host: pve14
- OSD: osd.41
- 2026-04-20 (Ceph 19.2.3)
- 2026-07-25 (Ceph 19.2.4)
Crash information
assert_condition: r == 0assert_file: ./src/os/bluestore/BlueStore.cc
assert_func: void BlueStore::_txc_apply_kv(TransContext*, bool)
assert_thread_name: bstore_kv_sync
stack_sig:
4bf8806194e49b144d67a8ec58abc1e8997fbbe795691d6d11b5dae2a4faa140
Backtrace:
BlueStore::_txc_apply_kv()
BlueStore::_kv_sync_thread()
BlueStore::KVSyncThread::entry()
The OSD aborts after the failed assertion.
Journal around the crash
Immediately before the assertion I only see normal RocksDB/BlueStore PutCF() operations:PutCF(...)
PutCF(...)
PutCF(...)
followed immediately by
FAILED ceph_assert(r == 0)
There are no preceding messages such as:
- RocksDB errors
- BlueFS errors
- Input/output errors
- ENOSPC
- checksum failures
- corruption messages
Observations
- The failure is identical on both 19.2.3 and 19.2.4.
- The stack signature is identical.
- The source line changed slightly between versions (14425 → 14445), which I assume is simply due to source changes.
- The crash occurs in the bstore_kv_sync thread while applying KV updates.
Questions
- Has anyone seen this assertion before?
- Is this a known Squid/BlueStore issue?
- Would you suspect:
- RocksDB/BlueStore metadata corruption,
- underlying storage problems,
- or a Ceph bug?
- Are there any recommended diagnostic steps beyond checking SMART/NVMe logs and kernel logs before considering repair or OSD recreation?
Thanks.