Test scripts: delete everything except chunk inventory test directory…

… and its content before hibernation and retire test to minimize test time. Shut down meta server before chunk servers. Do not fail recovery test if changing core file size limit does not succeed.

src/test-scripts/qfstest.sh

@@ -28,12 +28,13 @@ trdverify=${trdverify-0}
 s3debug=0
 jerasuretest=''
 mycsdebugverifyiobuffers=0
+myvalgrindlog='valgrind.log'
 
 while [ $# -ge 1 ]; do
     if [ x"$1" = x'-valgrind' ]; then
         myvalgrind='valgrind'
         myvalgrind="$myvalgrind"' -v'
-        myvalgrind="$myvalgrind"' --log-file=valgrind.log'
+        myvalgrind="$myvalgrind"' --log-file='"$myvalgrindlog"
         myvalgrind="$myvalgrind"' --leak-check=full'
         myvalgrind="$myvalgrind"' --leak-resolution=high'
         myvalgrind="$myvalgrind"' --show-reachable=yes'
@@ -1165,11 +1166,16 @@ echo "Testing chunk server hibernate and retire"
 # in the dumpster with replication larger than the number of chunk server
 # would prevent chunk server "retirement", until expiration / cleanup.
 # until they expire.
-for d in /user /sort; do
-    if runqfsroot -test -e "$d"; then
-        runqfsroot -rmr -skipTrash "$d" || exit
-    fi
-done
+
+rootrmlist=`runqfsroot -ls '/' \
+| awk '/^[d-]/{ if ($NF != "/dumpster" &&
+    $NF != "'"$chunkinventorytestdir"'") print $NF; }'`
+if [ x"$rootrmlist" = x ]; then
+    true
+else
+    runqfsroot -rmr -skipTrash $rootrmlist || exit
+fi
+
 movefromdumpster='/movefromdumpster.tmp'
 runqfsroot -mkdir "$movefromdumpster" || exit
 runqfsroot -ls '/dumpster' \
@@ -1188,7 +1194,6 @@ done || exit
 metaserversetparameter 'metaServer.panicOnRemoveFromPlacement=0'
 
 upserverslist='upservers.tmp'
-QFS_CLIENT_CONFIG= \
 runqfsadmin upservers > "$upserverslist" || exit
 
 # Tell chunk servers to re-connect to the meta server, in order to exercise
@@ -1241,6 +1246,9 @@ while [ $i -lt $e ]; do
         true
     else
         echo "Restarting chunk server $i"
+        if [ -e "$myvalgrindlog" ]; then
+            mv "$myvalgrindlog" "$myvalgrindlog"'.run.log' || exit
+        fi
         myrunprog "$chunkbindir"/chunkserver \
             "$chunksrvprop" "$chunksrvlog" >> "${chunksrvout}" 2>&1 &
         echo $! > "$chunksrvpid"
@@ -1250,10 +1258,14 @@ done
 cd "$testdir" || exit
 waitrecoveryperiodend
 
-# For now pause to let chunk server IOs complete
+# Allow meta server to run re-balancer
 sleep 5
 echo "Shutting down"
-pids=`getpids`
+kill -QUIT "$metapid" || exit
+
+pids=`getpids | grep -v "$metapid"`
+# For now pause to let chunk server IOs complete
+sleep 2
 for pid in $pids; do
     kill -QUIT "$pid" || exit
 done

src/test-scripts/recoverytest.sh

@@ -25,7 +25,7 @@
 # recovery by deleting chunk files and running file verification, and
 # using admin tool to force recovery of existing chunks.
 
-ulimit -c unlimited || exit
+ulimit -c unlimited
 
 builddir=`pwd`
 toolsdir=${toolsdir-"$builddir"/src/cc/tools}