[LU-8178] recovery_duration value becomes negative

Tue, 14 Jun 2016 23:02:47 +0000

Starting mds1: -o loop,user_xattr -o recovery_time_hard=60,recovery_time_soft=60  /tmp/lustre-mdt1 /mnt/mds1
Started lustre-MDT0000
recovery status
status: COMPLETE
recovery_start: 1385649466
recovery_duration: 18446744072323902150
completed_clients: 2/2
replayed_requests: 8737
last_transno: 8589950593
VBR: ENABLED
IR: DISABLED

Bug is occurred when /proc/fs/lustre/mdt/lustre-MDT0000/recovery_status is accessed after obd->obd_abort_recovery set to 0 and before target_finish_recovery call:

target_recovery_thread: obd->obd_recovering = obd->obd_abort_recovery = 0;
lprocfs_obd_rd_recovery_status: print("recovery_duration :", (obd->obd_recovery_end - obd->obd_recovery_start))
target_recovery_thread->target_finish_recovery: obd->obd_recovery_end = cfs_time_current_sec();

It seems bug exists for a long time.

Whamcloud Community JIRA

[LU-8178] recovery_duration value becomes negative