LU-6867 test: detect active facet based on current state

Lustre failover tests can not be ran test-by-test
on the setup with ${facet}_HOST != ${facet}failover_HOST
because of t-f does not restore facet state.
t-f keeps this info in "${facet}active" files, which are created
when facet_failover() is executed first time in the test session.
Before facet_failover() executed these files are empty and
active facet is ${facet} by default.
In case when tests are executed test-by-test the active facet is
${facet}failover after 1st test completed, and 2nd test is started
having ${facet}failover active without this info stored in
${facet}active files.

Patch contains the following changes:
- add the active facet detection based on current lustre state;
- fix sanity-hsm defect: exist with error if agt${n}1_HOST is empty.

Signed-off-by: Elena Gryaznova <elena.gryaznova@seagate.com>
Xyratex-bug-id: MRP-2680
Reviewed-by: Alexander Lezhoev <alexander.lezhoev@seagate.com>
Reviewed-by: Andriy Skulysh <andriy.skulysh@seagate.com>
Change-Id: Ie42baaa55a6433596e6004d16eb5c18ae2ef7479
Reviewed-on: https://review.whamcloud.com/15638
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Alexander Lezhoev <c17454@cray.com>
Reviewed-by: Andriy Skulysh <c17819@cray.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

diff --git a/lustre/tests/sanity-hsm.sh b/lustre/tests/sanity-hsm.sh
index 8669ba9..5a206b0 100755 (executable)
--- a/lustre/tests/sanity-hsm.sh
+++ b/lustre/tests/sanity-hsm.sh
@@ -100,6 +100,8 @@ init_agt_vars() {
                                agent=CLIENT2
                fi
                eval export agt${n}_HOST=\$\{agt${n}_HOST:-${!agent}\}
+               local var=agt${n}_HOST
+               [[ ! -z "${!var}" ]] || error "agt${n}_HOST is empty!"
        done
 
        export SINGLEAGT=${SINGLEAGT:-agt1}

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index bf9336a..1f53fc9 100755 (executable)
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -3215,20 +3215,59 @@ facet_failover_host() {
        fi
 }
 
+detect_active() {
+       local facet=$1
+       [ "$CLIENTONLY" ] && echo $facet && return
+
+       local failover=$(facet_failover_host $facet)
+
+       # failover is not associated with all facet types:
+       # "AGT" facet type (remote HSM agents) does not
+       # have a failover.
+       [[ -z "$failover" ]] && echo $facet && return
+
+       local host=$(facet_host $facet)
+       local dev=$(facet_device $facet)
+
+       # ${facet}_svc can not be used here because of
+       # facet_active() is called before this var initialized
+       local svc=$(do_node $host $E2LABEL ${dev})
+
+       # active facet is ${facet}failover if device is mounted on failover
+       # on other cases active facet is $facet
+       [[ $dev = $(do_node $failover \
+                       lctl get_param -n *.$svc.mntdev 2>/dev/null) ]] &&
+               echo ${facet}failover && return
+
+       echo $facet
+}
+
+init_active() {
+       local facet=$1
+
+       local active=$(detect_active $facet)
+       echo "${facet}active=$active" > $TMP/${facet}active
+}
+
 facet_active() {
-    local facet=$1
-    local activevar=${facet}active
+       local facet=$1
+       local activevar=${facet}active
 
-    if [ -f $TMP/${facet}active ] ; then
-        source $TMP/${facet}active
-    fi
+       # file is missing (nothing to store) if fail() is not
+       # executed during this test session yet;
+       # file content:
+       #      ost1active=ost1failover
+       #      ost1active=ost1
+       # let's detect active facet based on current lustre state
+       if [ ! -f $TMP/${facet}active ] ; then
+               init_active $facet
+       fi
+       source $TMP/${facet}active
 
-    active=${!activevar}
-    if [ -z "$active" ] ; then
-        echo -n ${facet}
-    else
-        echo -n ${active}
-    fi
+       # is ${facet}active set somewhere else?
+       active=${!activevar}
+       [[ -z "$active" ]] && exit 1
+       echo -n ${active}
 }
 
 facet_active_host() {
@@ -3714,13 +3753,6 @@ unmount_fstype() {
 ## MountConf setup
 
 stopall() {
-    # make sure we are using the primary server, so test-framework will
-    # be able to clean up properly.
-    activemds=`facet_active mds1`
-    if [ $activemds != "mds1" ]; then
-        fail mds1
-    fi
-
     local clients=$CLIENTS
     [ -z $clients ] && clients=$(hostname)