[LU-13450] handle corrupted llog files

• Type: Bug
• Resolution: Unresolved
• Priority: Minor
• Affects Version/s: Lustre 2.14.0
• Severity: 3
• Assignee: Alex Zhuravlev
• Reporter: Alex Zhuravlev

Description

if plain llog got corrupted for a reason, then MDS crashes:

LustreError: 9838:0:(llog_osd.c:259:llog_osd_read_header()) lustre-MDT0000-osd: bad log [0x1:0x6:0x0] header magic: 0x0 (expected 0x10645539)
Lustre: 9838:0:(llog_cat.c:830:llog_cat_process_common()) lustre-OST0001-osc-MDT0000: can't find llog handle [0x6:0x1:0x0]:0: rc = -5
LustreError: 9838:0:(llog.c:730:llog_process_thread()) lustre-OST0001-osc-MDT0000: Local llog found corrupted #0x5:1:0 catalog index 1 count 3
LustreError: 9838:0:(osp_sync.c:1284:osp_sync_thread()) ASSERTION( kthread_should_stop() ) failed: 0 changes, 0 in progress, 0 in flight
LustreError: 9838:0:(osp_sync.c:1284:osp_sync_thread()) LBUG

{coe}

the above is reproduced on master branch, the test is scripted and will be included into the patch.