Whamcloud - gitweb
LU-10937 mgc: restore mgc binding for sptlrpc 11/33311/7
authorJames Simmons <uja.ornl@yahoo.com>
Wed, 7 Nov 2018 14:08:30 +0000 (09:08 -0500)
committerOleg Drokin <green@whamcloud.com>
Tue, 13 Nov 2018 06:16:08 +0000 (06:16 +0000)
commitca9300e53dc2b7bcaaa5482bb4234cce7d9a344e
treec1fc033b39bde4c95212d7d08354750a86769ee0
parente2bd32ca75d60870a70cd3a00e8aac8efb751762
LU-10937 mgc: restore mgc binding for sptlrpc

The work for LU-9034 mapped config logs to separate mgc devices.
This change prevented the ability to configure sptlrpc. A later
work around was introduced in LU-9567. Recently it was reported
that the work around introduced can now cause a MGC failover
panic. This patch is the proper fix in that the sptlrpc is
properly bound to an mgc device.

The sptlrpc config record expects 2 pieces of data:

  *  [0]: fs_name/target_name,
  *  [1]: rule string

What was happening is that when you set cfg_instance it was used
to create a new instance name of the form fsname-%p. For sptlrpc
it expects it to only be fsname. The solution is to test if the
config record is for sptlrpc and in that can keep the first
record field as is. With this change we can drop cfg_obdname
which only sptlrpc used.

Test-Parameters: testlist=sanity-gss envdefinitions=ONLY=1,SHARED_KEY=true
Test-Parameters: testlist=sanity-sec envdefinitions=SHARED_KEY=true

Change-Id: I785f98264c6269f95c0d9a564b731d1b6ff0bcee
Signed-off-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-on: https://review.whamcloud.com/33311
Tested-by: Jenkins
Reviewed-by: Sebastien Buisson <sbuisson@ddn.com>
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
lustre/include/obd_class.h
lustre/mgc/mgc_request.c
lustre/obdclass/obd_config.c