从 sql-client.sh 提交的 Flink 作业，如何从保存点恢复？

Question

I submitted a job from apache-flink sql-client, and created a savepoint. The issue is the meta data doesn't contain jar and classname, let alone the parameters. How to restart?

Answer 1

flink run -s hdfs://ns/flink/flink-checkpoints/savepoint-c5dade-af74904ab30c -m yarn-cluster -yid application_1539849585041_0459 -c org.apache.flink.table.client.SqlClient opt/flink-sql-client-1.6.1.jar embedded -e /opt/flink/flink-bin/yarn-app/a/sql-client-kafka-json.yaml --library /opt/flink/flink-bin//lib -u 'INSERT INTO testjsonSink SELECT * FROM testjsonSource;'`

Adjust jars lib. It been solved as before. Poor documentation.

Answer 2

official working with 1.12.1 with scala 1.12: flink run -s hdfs://dbt1caw005.webex.com:9000/flink-checkpoints/savepoint-dafd7c-05d66b098493 -C file:///opt/flink/jars/flink-python_2.12-1.12.1.jar -c org.apache.flink.table.client.SqlClient /opt/flink/opt/flink-sql-client_2.12-1.12.1.jar embedded -e /vdb/sql.yml -l /opt/flink/jars -u "INSERT INTO CALL_DURATION_USER SELECT orgId, userId, window_start, window_end, total_minutes, total_calls FROM ( SELECT , ROW_NUMBER() OVER (PARTITION BY orgId,window_end ORDER BY total_minutes desc) AS rownum FROM ( SELECT orgId, userId, HOP_START(ts, INTERVAL '1' DAY,INTERVAL '30' DAY) window_start, HOP_END(ts, INTERVAL '1' DAY,INTERVAL '30' DAY) window_end, CAST(sum(cast(legDuration as bigint)/60 ) AS BIGINT) total_minutes, CAST(count( ) AS BIGINT) total_calls FROM callduration_ts GROUP BY HOP(ts, INTERVAL '1' DAY,INTERVAL '30' DAY),orgId, userId ) ) WHERE rownum < 101"