Details
-
Bug
-
Resolution: Not a Bug
-
Critical
-
None
-
Lustre 2.10.4
-
None
-
3
-
9223372036854775807
Description
I have noticed that from 2.10.3 to b2_10 MGS startup is happening asynchronously and taking much longer.
In 2.10.3, immediately after (or within a very small number of seconds at most) the mount for the MGS returns it reports as being up:
# mount -t lustre zfs_pool_scsi0QEMU_QEMU_HARDDISK_disk13/MGS /mnt/MGS # lctl dl: 0 UP osd-zfs MGS-osd MGS-osd_UUID 4 1 UP mgs MGS MGS 4 2 UP mgc MGC10.14.83.84@tcp f2845b74-6eb5-07a5-1aaa-d7c54f76df5f 4
However the same operation on b2_10 is taking on the order of 10-11 seconds after the mount returns before lctl dl shows the MGS as being up.
I have attached debug-lctl-1519949571.log.xz
which is a lustre debug log of the mount an lctl conf_param where it takes 10-11 seconds before the lctl conf_param returns with success instead of a -19.