From: Xiaolin (Charlene) Zang <xiaolinzang@microsoft.com>
Date: Thu, 8 Jul 2021 04:32:40 +0000 (-0400)
Subject: LU-17727 tests: add to auster --stop-on-error option
X-Git-Tag: 2.15.63~5
X-Git-Url: https://git.whamcloud.com/?a=commitdiff_plain;h=refs%2Fchanges%2F55%2F54755%2F3;p=fs%2Flustre-release.git

LU-17727 tests: add to auster --stop-on-error option

add to auster --stop-on-error option, a comma separated list of tests.

If any such test fails, auster will exit immediately without any
cleanup to make debugging particularly difficult and rare bugs more
tractable.

Signed-off-by: Xiaolin (Charlene) Zang <xiaolinzang@microsoft.com>
Change-Id: Icd8d1eaf8ae799bd74f9147ac9080a0950977526
Reviewed-on: https://review.whamcloud.com/c/fs/lustre-release/+/54755
Tested-by: jenkins <devops@whamcloud.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Reviewed-by: Charlie Olmstead <charlie@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>
---

diff --git a/lustre/tests/auster b/lustre/tests/auster
index 4fefde8..b352d9f 100755
--- a/lustre/tests/auster
+++ b/lustre/tests/auster
@@ -165,6 +165,10 @@ if $client_tests_only; then
 	export OSTCOUNT="$(( $($LFS osts | wc -l) - 1 ))"
 fi
 
+# special return code to stop testing without cleanup
+# used with the --stop-on-error option
+export STOP_NOW_RC=111
+
 if [ $upload_logs = true ] ; then
     upload_script=$(find_script_in_path maloo_upload.sh $PATH:$LUSTRE/tests)
     if [[ -z $upload_script ]]; then
diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index 15bf137..aac78ee 100755
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -252,6 +252,11 @@ run_suite() {
 	log_test_status $duration $status
 	[[ ! -f $TF_SKIP ]] || rm -f $TF_SKIP
 
+	# got STOP_NOW_RC, return immediately before reset
+	[[ $rc -eq $STOP_NOW_RC ]] &&
+		echo "stop testing on rc $STOP_NOW_RC" &&
+		return $STOP_NOW_RC
+
 	reset_lustre
 
 	return $rc
@@ -343,6 +348,11 @@ run_suites() {
 					export STOP_AT=$1
 
 					opts+="STOP_AT=$STOP_AT ";;
+				--stop-on-error)
+					shift;
+					export STOP_ON_ERROR=$(split_commas $1)
+
+					opts+="STOP_ON_ERROR=$STOP_ON_ERROR ";;
 				--time-limit)
 					shift;
 					time_limit=$1;;
@@ -358,6 +368,9 @@ run_suites() {
 			run_suite_logged $suite || RC=$?
 			unset first_suite
 			echo $suite returned $RC
+
+			# stop testing immediately if rc is STOP_NOW_RC
+			[[ $RC -eq $STOP_NOW_RC ]] && exit $STOP_NOW_RC
 		fi
 		done
 	if $upload_logs; then
@@ -7144,6 +7157,11 @@ build_test_filter() {
 	for G in ${GRANT_CHECK_LIST//[+,]/ }; do
 		eval GCHECK_ONLY_${G}=true
 	done
+	# similar to $EXCEPT, STOP_ON_ERROR is a list of test numbers,
+	# e.g. [30d, 34a].  Now set variable STOP_ON_ERROR_30d, etc.
+	for T in $STOP_ON_ERROR; do
+		eval STOP_ON_ERROR_${T}=true
+	done
 }
 
 basetest() {
@@ -7441,6 +7459,12 @@ run_one_logged() {
 		fi
 
 		log_sub_test_end $TEST_STATUS $duration_sub "$rc" "$test_error"
+
+		# exit test suite if the failed test is in STOP_ON_ERROR list
+		[[ $TEST_STATUS == "FAIL" ]] &&
+			[[ -v STOP_ON_ERROR_$testnum ]] &&
+			exit $STOP_NOW_RC
+
 		[[ $rc != 0 || "$TEST_STATUS" != "PASS" ]] && break
 	done