[LU-10379] "lfs changelog" generates constant "Changelog: Input/output error" errors

• Type: Bug
• Resolution: Duplicate
• Priority: Critical
• Fix Version/s: Lustre 2.11.0
• Affects Version/s: Lustre 2.11.0
• Severity: 3
• Assignee: WC Triage
• Reporter: Andreas Dilger

Description

I'm seeing lots of errors from "lfs changelog" and "lfs changelog_clear" commands:
https://testing.hpdd.intel.com/test_sets/70f5331a-db46-11e7-a066-52540065bddc

verifying changelog mask
CMD: trevis-13vm4 /usr/sbin/lctl set_param mdd.*.changelog_mask=-MKDIR
CMD: trevis-13vm4 /usr/sbin/lctl set_param mdd.*.changelog_mask=-CLOSE
CMD: trevis-13vm4 /usr/sbin/lctl set_param mdd.*.changelog_mask=+MKDIR
CMD: trevis-13vm4 /usr/sbin/lctl set_param mdd.*.changelog_mask=+CLOSE
Changelog: Input/output error
lustre-MDT0000.7 01CREAT 11:20:48.951746235 2017.12.07 0x0 t=[0x200002b12:0x89c:0x0] j=cp.0 ef=0x0 p=[0x200002b12:0x89a:0x0] pic1.jpg
lustre-MDT0000.8 11CLOSE 11:20:48.955294645 2017.12.07 0xc2 t=[0x200002b12:0x89c:0x0] j=cp.0 ef=0x0
:
Changelog: Input/output error
Changelog: Input/output error
verifying target fid
Changelog: Input/output error
verifying parent fid
Changelog: Input/output error
getting records for cl1

This doesn't appear to be affecting functionality, but makes these commands difficult to use.

Issue Links

• is related to: LU-10218 errors not propagated to changelog readers
• is related to: LU-10267 Wrong poll() returned revents for changelog device