[LU-7662] lfsck don't complete

• Type: Bug
• Resolution: Duplicate
• Priority: Blocker
• Fix Version/s: Lustre 2.8.0
• Labels: soak
• Environment: lola; build: master, 2.7.64-81-g6fc8da4, 6fc8da41f2ff5156639e89f379adcdbb73ac8567
• Severity: 3
• Assignee: nasf (Inactive)
• Reporter: Frank Heckes (Inactive)

Description

Error happened during lfsck run of soak FS using build '20160108'. (see https://wiki.hpdd.intel.com/display/Releases/Soak+Testing+on+Lola#SoakTestingonLola-20160108)
DNE is enabled.

• lfsck started on MDS hosting mdt-0:

  [root@lola-8 ~]# date; lctl lfsck_start -M soaked-MDT0000 -s 1000 -t all -A ; date
  Wed Jan 13 04:42:28 PST 2016
  Started LFSCK on the device soaked-MDT0000: scrub layout namespace
  Wed Jan 13 04:42:28 PST 2016
  

  No soak test was running

• lfsck_namespace don't complete phase scanning-phase2
• MDSes lola-9,11 showed an increasing number of blocked mdt_out* - threads
• Triggering stack trace lead kernel panic on lola-11 (2016-01-13-08:15:22)
• All MDSes show only minimal utilization of system resources

Attached files:

• console, messages files of lola-9,11; containing stack trace information
• vmcore-dmesg.txt of lola-11
• lfsck status information of all MDTs

Crash file location see next comment.

Issue Links

• duplicates: LU-6684 lctl lfsck_stop hangs