[LU-9400] conf-sanity test 103 hangs with separate MGS and MDS

Related — Mon, 17 Jun 2019 15:57:38 +0000

conf-sanity test_103 hangs when the MGS and MDS are on separate nodes.

Looking at the test_renamfs() routine in conf-sanity.sh, the first thing it does is check if we have a combined MDS and MGS. If we do, then rename the MGS. After that, we rename the MGS:

7197 test_renamefs() {
7198         local newname=$1
7199 
7200         echo "rename $FSNAME to $newname"
7201 
7202         if [ ! combined_mgs_mds ]; then
7203                 local facet=$(mgsdevname)
7204 
7205                 do_facet mgs \
7206                         "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7207                         error "(7) Fail to rename MGS"
7208                 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7209                         reimport_zpool mgs $newname-mgs
7210                 fi
7211         fi
7212 
7213         for num in $(seq $MDSCOUNT); do
7214                 local facet=$(mdsdevname $num)
7215 
7216                 do_facet mds${num} \
7217                         "$TUNEFS --fsname=$newname --rename=$FSNAME -v $facet"||
7218                         error "(8) Fail to rename MDT $num"
7219                 if [ "$(facet_fstype $facet)" = "zfs" ]; then
7220                         reimport_zpool mds${num} $newname-mdt${num}
7221                 fi
7222         done
…

Yet, looking at the console output, we see that we enter the test_renamefs() routine and, with a separate MDS/MGS, we just skip over renaming the MGS and rename the MDS:

rename scratch to mylustre
checking for existing Lustre data: found
Reading CONFIGS/mountdata

   Read previous values:
Target:     scratch-MDT0000
Index:      0
Lustre FS:  scratch
Mount type: ldiskfs
Flags:      0x1
              (MDT )
Persistent mount opts: user_xattr,errors=remount-ro
Parameters: mgsnode=10.100.4.154@tcp sys.timeout=20 lov.stripesize=1048576 lov.stripecount=0 mdt.identity_upcall=/usr/sbin/l_getidentity
…

The problem is that the test_renamefs() routine checks for a combined MDS/MGS (calls combined_mgs_mds) in the “test” brackets and test doesn’t execute the combined_mgs_mds routine.

The fix is to remove the “[“ “]” from line 7202 of conf-sanity.sh.

Whamcloud Community JIRA

[LU-9400] conf-sanity test 103 hangs with separate MGS and MDS