deep fsck fails on inspecting very large onodes
Steps to reproduce (100%):
Put 3GB object to replicated pool via rados then do deep fsck either on mount or via ceph-bluestore-tool.
Following output is provided:
2019-01-28 17:17:55.960 7f5d42af2b40 -1 bdev(0x557b43254700 dev/osd0/block) aio_submit retries 5423
2019-01-28 17:17:57.836 7f5d42af2b40 -1 bluestore(dev/osd0) fsck error: #4:238134c4:::tmp_3g:head# error during read: (1073741824) Unknown error 1073741824
Looks like aio_queue_t::submit_batch timeouts due to long list of blocks to read. We need to cap this for fsck and, probably, introduce a sort of throttling for regular reads...
- Subject changed from deep fsck fails on inspecting objects >2Gb length to deep fsck fails on inspecting objects very large onodes
- Subject changed from deep fsck fails on inspecting objects very large onodes to deep fsck fails on inspecting very large onodes
- Status changed from New to Need Review
- Backport set to mimic,luminous
- Status changed from Need Review to Pending Backport
- Copied to Backport #38187: mimic: deep fsck fails on inspecting very large onodes added
- Copied to Backport #38188: luminous: deep fsck fails on inspecting very large onodes added
- Status changed from Pending Backport to Resolved
Also available in: Atom