LU-1538 tests: clarify conf-sanity test_49 failures

Clean up the code for conf-sanity.sh test_49 to allow easier debugging
of test failures, which were seen intermittently during development.

Also clean up the coding style to match current guidelines.

Signed-off-by: Andreas Dilger <andreas.dilger@intel.com>
Change-Id: Iac60fd89c5f193bfa11d9223e75ae947c8b6e186
Reviewed-on: http://review.whamcloud.com/6207
Tested-by: Jenkins
Tested-by: Maloo <hpdd-maloo@intel.com>
Reviewed-by: Bob Glossman <bob.glossman@intel.com>
Reviewed-by: Jian Yu <jian.yu@intel.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>

diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh
index 93cdc9c..af12da1 100644 (file)
--- a/lustre/tests/conf-sanity.sh
+++ b/lustre/tests/conf-sanity.sh
@@ -1,9 +1,9 @@
 #!/bin/bash
 
-# FIXME - there is no reason to use all of these different
-#   return codes, espcially when most of them are mapped to something
-#   else anyway.  The combination of test number and return code
-#   figure out what failed.
+# FIXME - there is no reason to use all of these different return codes,
+#   espcially when most of them are mapped to something else anyway.
+#   The tests should use error() to describe the failure more clearly,
+#   and reduce the need to look into the tests to see what failed.
 
 set -e
 
@@ -2726,7 +2726,7 @@ test_48() { # bug 17636
 run_test 48 "too many acls on file"
 
 # check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE
-test_49() { # bug 17710
+test_49a() { # bug 17710
        local timeout_orig=$TIMEOUT
        local ldlm_timeout_orig=$LDLM_TIMEOUT
        local LOCAL_TIMEOUT=20
@@ -2736,49 +2736,60 @@ test_49() { # bug 17710
 
        reformat
        setup_noconfig
-       check_mount || return 1
+       check_mount || error "client mount failed"
 
        echo "check ldlm_timout..."
-       LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
-       LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
-       LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
+       local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
+       local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
+       local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
 
-       if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
+       if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
                error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
        fi
 
        if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT / 3)) ]; then
-               error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
+               error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT / 3))"
        fi
 
        umount_client $MOUNT
-       stop_ost || return 2
-       stop_mds || return 3
+       stop_ost || error "problem stopping OSS"
+       stop_mds || error "problem stopping MDS"
+
+       LDLM_TIMEOUT=$ldlm_timeout_orig
+       TIMEOUT=$timeout_orig
+}
+run_test 49a "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
+
+test_49b() { # bug 17710
+       local timeout_orig=$TIMEOUT
+       local ldlm_timeout_orig=$LDLM_TIMEOUT
+       local LOCAL_TIMEOUT=20
 
        LDLM_TIMEOUT=$((LOCAL_TIMEOUT - 1))
+       TIMEOUT=$LOCAL_TIMEOUT
 
        reformat
        setup_noconfig
-       check_mount || return 7
+       check_mount || error "client mount failed"
 
-       LDLM_MDS="`do_facet $SINGLEMDS lctl get_param -n ldlm_timeout`"
-       LDLM_OST1="`do_facet ost1 lctl get_param -n ldlm_timeout`"
-       LDLM_CLIENT="`do_facet client lctl get_param -n ldlm_timeout`"
+       local LDLM_MDS="$(do_facet $SINGLEMDS lctl get_param -n ldlm_timeout)"
+       local LDLM_OST1="$(do_facet ost1 lctl get_param -n ldlm_timeout)"
+       local LDLM_CLIENT="$(do_facet client lctl get_param -n ldlm_timeout)"
 
-       if [ $LDLM_MDS -ne $LDLM_OST1 ] || [ $LDLM_MDS -ne $LDLM_CLIENT ]; then
+       if [ $LDLM_MDS -ne $LDLM_OST1 -o $LDLM_MDS -ne $LDLM_CLIENT ]; then
                error "Different LDLM_TIMEOUT:$LDLM_MDS $LDLM_OST1 $LDLM_CLIENT"
        fi
 
        if [ $LDLM_MDS -ne $((LOCAL_TIMEOUT - 1)) ]; then
-               error "LDLM_TIMEOUT($LDLM_MDS) is not correct"
+               error "LDLM_TIMEOUT($LDLM_MDS) is not $((LOCAL_TIMEOUT - 1))"
        fi
 
-       cleanup || return $?
+       cleanup || error "cleanup failed"
 
        LDLM_TIMEOUT=$ldlm_timeout_orig
        TIMEOUT=$timeout_orig
 }
-run_test 49 "check PARAM_SYS_LDLM_TIMEOUT option of MKFS.LUSTRE"
+run_test 49b "check PARAM_SYS_LDLM_TIMEOUT option of mkfs.lustre"
 
 lazystatfs() {
         # Test both statfs and lfs df and fail if either one fails