LU-10632 tests: recovery-small test_26 idle_timeout

In recovery-small test_26() use "lfs df" instead of plain "df"
since statfs may be fetched from the MDS cache and will not
ensure that the client->OST connections are currently active.

Also, check a few entries further back in the OSC state log for an
EVICTED message, in case the client idle disconnects from the server
again while checking all of the imports.

Test-Parameters: trivial testlist=recovery-small env=ONLY=26a,ONLY_REPEAT=100
Fixes: 5a6ceb664f07 ("LU-7236 ptlrpc: idle connections can disconnect")
Signed-off-by: Andreas Dilger <adilger@whamcloud.com>
Change-Id: I8c370cb75f4e06258ef3c032630fc20354a15dcc
Reviewed-on: https://review.whamcloud.com/42006
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
Reviewed-by: James Nunez <jnunez@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

diff --git a/lustre/tests/recovery-small.sh b/lustre/tests/recovery-small.sh
index 8b4b594..77992a7 100755 (executable)
--- a/lustre/tests/recovery-small.sh
+++ b/lustre/tests/recovery-small.sh
@@ -1069,6 +1069,8 @@ test_26a() {      # was test_26 bug 5921 - evict dead exports by pinger
 
        check_timeout || return 1
 
+       # make sure all imports are connected and not IDLE
+       do_facet client lfs df > /dev/null
 # OBD_FAIL_PTLRPC_DROP_RPC 0x505
        do_facet client lctl set_param fail_loc=0x505
        local before=$(date +%s)
@@ -1079,7 +1081,7 @@ test_26a() {      # was test_26 bug 5921 - evict dead exports by pinger
        # the loser might have to wait for the next ping.
        sleep $((TIMEOUT * 2 + TIMEOUT * 3 / 4))
        do_facet client lctl set_param fail_loc=0x0
-       do_facet client df > /dev/null
+       do_facet client lfs df > /dev/null
 
        local oscs=$(lctl dl | awk '/-osc-/ {print $4}')
        check_clients_evicted $before ${oscs[@]}
@@ -1117,10 +1119,9 @@ test_26b() {      # bug 10140 - evict dead exports by pinger
        # = 9 * PING_INTERVAL + PING_INTERVAL
        # = 10 PING_INTERVAL = 10 obd_timeout / 4 = 2.5 obd_timeout
        # let's wait $((TIMEOUT * 3)) # bug 19887
-       local rc=0
-       wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 3)) || \
-               error "Client was not evicted by ost" rc=1
-       wait_client_evicted $SINGLEMDS $MDS_NEXP $((TIMEOUT * 3)) || \
+       wait_client_evicted ost1 $OST_NEXP $((TIMEOUT * 3)) ||
+               error "Client was not evicted by ost"
+       wait_client_evicted $SINGLEMDS $MDS_NEXP $((TIMEOUT * 3)) ||
                error "Client was not evicted by mds"
 }
 run_test 26b "evict dead exports"

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 46601ad..71d4aa3 100755 (executable)
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -9572,14 +9572,17 @@ check_clients_evicted() {
        local rc=0
 
        for osc in $oscs; do
-               ((rc++))
                echo "Check state for $osc"
                local evicted=$(do_facet client $LCTL get_param osc.$osc.state |
-                       tail -n 3 | awk -F"[ [,]" \
-                       '/EVICTED ]$/ { if (mx<$5) {mx=$5;} } END { print mx }')
+                       tail -n 5 | awk -F"[ ,]" \
+                       '/EVICTED/ { if (mx<$4) { mx=$4; } } END { print mx }')
                if (($? == 0)) && (($evicted > $before)); then
                        echo "$osc is evicted at $evicted"
-                       ((rc--))
+               else
+                       ((rc++))
+                       echo "$osc was not evicted after $before:"
+                       do_facet client $LCTL get_param osc.$osc.state |
+                               tail -n 8
                fi
        done