[CORE-7961] rptest: Add ducktape test for partition movement in RRR cluster #24159
Conversation
Signed-off-by: Evgeny Lazin <[email protected]>
|
the below tests from https://buildkite.com/redpanda/redpanda/builds/58193#01933fe3-0d52-4e33-9640-662e140cfed0 have failed and will be retried the below tests from https://buildkite.com/redpanda/redpanda/builds/58697#019363ec-d093-4aea-b1e7-a31e5d1ea5f6 have failed and will be retried |
|
non flaky failures in https://buildkite.com/redpanda/redpanda/builds/58193#01934040-98ba-4bc6-81b0-58fdd6bca7ce: |
Retry command for Build#58193please wait until all jobs are finished before running the slash command |
| for part_id in range(0, partition_count): | ||
| assignments = self._get_node_assignments(admin, self.topic_name, | ||
| part_id) | ||
| self.logger.info( | ||
| f"initial assignments for {self.topic_name}/{part_id}: {assignments}" | ||
| ) | ||
| replicas = set([r['node_id'] for r in assignments]) | ||
| for b in brokers: | ||
| if b['node_id'] not in replicas: | ||
| assignments[0] = {"node_id": b['node_id']} | ||
| break | ||
| self.logger.info( | ||
| f"new assignments for {self.topic_name}/{part_id}: {assignments}" | ||
| ) | ||
| self._set_partition_assignments(self.topic_name, part_id, | ||
| assignments, admin) |
There was a problem hiding this comment.
is the goal to run partition movement concurrently with data arriving into the destination cluster?
There was a problem hiding this comment.
not necessary, just to make sure that fast partition movement is not broken for RRR
we have a gap in our testing here
There was a problem hiding this comment.
should we verify that fast partition movement was used?
|
/ci-repeat 1 |
| self.start_consumer() | ||
|
|
||
| # Wait until reconfiguration are completed and start consuming again | ||
| wait_until(lambda: len(admin.list_reconfigurations()) == 0, 30) |
There was a problem hiding this comment.
please add an informative error message parameter to this wait_until
| wait_until( | ||
| lambda: len(admin.list_reconfigurations()) == partition_count, 30) |
There was a problem hiding this comment.
please add an informative error message parameter to this wait_until
|
|
||
| # Wait until reconfiguration are completed and start consuming again | ||
| wait_until(lambda: len(admin.list_reconfigurations()) == 0, 30) | ||
| self.start_consumer() |
There was a problem hiding this comment.
why do we start the consumer and then immediately end the test?
| for part_id in range(0, partition_count): | ||
| assignments = self._get_node_assignments(admin, self.topic_name, | ||
| part_id) | ||
| self.logger.info( | ||
| f"initial assignments for {self.topic_name}/{part_id}: {assignments}" | ||
| ) | ||
| replicas = set([r['node_id'] for r in assignments]) | ||
| for b in brokers: | ||
| if b['node_id'] not in replicas: | ||
| assignments[0] = {"node_id": b['node_id']} | ||
| break | ||
| self.logger.info( | ||
| f"new assignments for {self.topic_name}/{part_id}: {assignments}" | ||
| ) | ||
| self._set_partition_assignments(self.topic_name, part_id, | ||
| assignments, admin) |
There was a problem hiding this comment.
should we verify that fast partition movement was used?
Add new ducktape test. The test creates two clusters (source and RRR), produces data to the source and consumes from the RRR. The test initiates partition movement for every partition of the RRR cluster.
Backports Required
Release Notes