Big thanks to Sergey Nuyanzin for his suggestion on the JVM thread dumps :)
On Thu, Jun 15, 2023 at 1:19 PM Martijn Visser <martijnvis...@apache.org> wrote: > Hi all, > > I would like to inform you of two changes that have been made to the > shared CI workflow that's used for Flink's externalized connectors. > > 1. Up until now, weekly builds were running to validate that connector > code (still) works with Flink. However, these builds were only running for > code on the "main" branch of the connector, and not for the branches of the > connector (like v3.0 for Elasticsearch, v1.0 for Opensearch etc). This was > tracked under https://issues.apache.org/jira/browse/FLINK-31923. > > That issue has now been fixed, with the Github Action workflow now > accepting a map with arrays, which can contain a combination of Flink > versions to test for and the connector branch it should test. See > https://github.com/apache/flink-connector-jdbc/blob/main/.github/workflows/weekly.yml#L28-L47 > for an example on the Flink JDBC connector > > This change has already been applied on the externalized connectors GCP > PubSub, RabbitMQ, JDBC, Pulsar, MongoDB, Opensearch, Cassandra, > Elasticsearch. AWS is pending the merging of the PR. For Kafka, Hive and > HBase, since they haven't finished externalization, this isn't applicable > to them yet. > > 2. When working on the debugging of a problem with the JDBC connector, one > of the things that was needed to debug that problem was the ability to see > the JVM thread dump. With > https://issues.apache.org/jira/browse/FLINK-32331 now completed, every > failed CI run will have a JVM thread dump. You can see the implementation > for that in > https://github.com/apache/flink-connector-shared-utils/blob/ci_utils/.github/workflows/ci.yml#L161-L195 > > Best regards, > > Martijn >