Github user dawidwys commented on a diff in the pull request:
https://github.com/apache/flink/pull/6415#discussion_r205080760
--- Diff: flink-end-to-end-tests/test-scripts/test_ha_dataset.sh ---
@@ -0,0 +1,139 @@
+#!/usr/bin/env bash
+################################################################################
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+################################################################################
+
+source "$(dirname "$0")"/common.sh
+
+TEST_PROGRAM_JAR=${END_TO_END_DIR}/flink-dataset-allround-test/target/DataSetAllroundTestProgram.jar
+
+JM_WATCHDOG_PID=0
+
+# flag indicating if we have already cleared up things after a test
+CLEARED=0
+
+function stop_cluster_and_watchdog() {
+ if [ ${CLEARED} -eq 0 ]; then
+
+ if ! [ ${JM_WATCHDOG_PID} -eq 0 ]; then
+ echo "Killing JM watchdog @ ${JM_WATCHDOG_PID}"
+ kill ${JM_WATCHDOG_PID} 2> /dev/null
+ wait ${JM_WATCHDOG_PID} 2> /dev/null
+ fi
+
+ CLEARED=1
+ fi
+}
+
+function verify_logs() {
+ local OUTPUT=$FLINK_DIR/log/*.out
+ local JM_FAILURES=$1
+ local EXIT_CODE=0
+
+ # verify that we have no alerts
+ if ! [ `cat ${OUTPUT} | wc -l` -eq 0 ]; then
+ echo "FAILURE: Alerts found at the general purpose DataSet job."
+ EXIT_CODE=1
+ fi
+
+ # checks that all apart from the first JM recover the failed jobgraph.
+ if ! [ `grep -r --include '*standalonesession*.log' 'Recovered
SubmittedJobGraph' "${FLINK_DIR}/log/" | cut -d ":" -f 1 | uniq | wc -l` -eq
${JM_FAILURES} ]; then
+ echo "FAILURE: A JM did not take over."
+ EXIT_CODE=1
+ fi
+
+ if [[ $EXIT_CODE != 0 ]]; then
+ echo "One or more tests FAILED."
+ exit $EXIT_CODE
+ fi
+}
+
+function jm_watchdog() {
+ local EXPECTED_JMS=$1
+ local IP_PORT=$2
+
+ while true; do
+ local RUNNING_JMS=`jps | grep 'StandaloneSessionClusterEntrypoint'
| wc -l`;
+ local MISSING_JMS=$((EXPECTED_JMS-RUNNING_JMS))
+ for (( c=0; c<MISSING_JMS; c++ )); do
+ "$FLINK_DIR"/bin/jobmanager.sh start "localhost" ${IP_PORT}
+ done
+ sleep 1;
+ done
+}
+
+function kill_jm {
+ local JM_PIDS=`jps | grep 'StandaloneSessionClusterEntrypoint' | cut
-d " " -f 1`
+ local JM_PIDS=(${JM_PIDS[@]})
+ local PID=${JM_PIDS[0]}
+ kill -9 ${PID}
+
+ echo "Killed JM @ ${PID}"
+}
+
+function run_ha_test() {
+ local PARALLELISM=$1
+
+ local JM_KILLS=3
+
+ CLEARED=0
+ mkdir -p ${TEST_DATA_DIR}/control
+ touch ${TEST_DATA_DIR}/control/test.txt
+
+ # start the cluster on HA mode
+ start_ha_cluster
+
+ echo "Running on HA mode: parallelism=${PARALLELISM}."
+
+ # submit a job in detached mode and let it run
+ local JOB_ID=$($FLINK_DIR/bin/flink run -d -p ${PARALLELISM} \
+ $TEST_PROGRAM_JAR \
+ --loadFactor 4 \
+ --outputPath $TEST_DATA_DIR/out/dataset_allround \
+ --source ${TEST_DATA_DIR}/control/test.txt \
+ | grep "Job has been submitted with JobID" | sed 's/.* //g')
+
+ wait_job_running ${JOB_ID}
+
+ # start the watchdog that keeps the number of JMs stable
+ jm_watchdog 1 "8081" &
+ JM_WATCHDOG_PID=$!
+ echo "Running JM watchdog @ ${JM_WATCHDOG_PID}"
+
+ for (( c=0; c<${JM_KILLS}; c++ )); do
+ # kill the JM and wait for watchdog to
+ # create a new one which will take over
+ kill_jm
+ sleep 20
+ wait_job_running ${JOB_ID}
+ done
+
+ printf "STOP" >> ${TEST_DATA_DIR}/control/test.txt
--- End diff --
We've verified that it started being recovered. We check for string
"Recovered SubmittedJobGraph" It might happen that it won't be able to run,
e.g. it might not be able to allocate resources.
I think with cancel we would not cancel e.g. this one:
https://issues.apache.org/jira/browse/FLINK-9635. Am I correct?
---
