-
Notifications
You must be signed in to change notification settings - Fork 553
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fixed large allocation in kafka::wait_for_leaders
#16287
Fixed large allocation in kafka::wait_for_leaders
#16287
Conversation
/dt |
/ci-repeat 10 |
/ci-repeat 10 |
1 similar comment
/ci-repeat 10 |
20666d7
to
204e1f5
Compare
/ci-repeat 10 |
ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/44441#018d5671-7d76-49d7-a337-d78544fb0815 ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/44441#018d5671-7d75-45c1-983f-293bda48ad87 ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/44441#018d5671-7d77-4000-99ae-f3f7f04f915f ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/44441#018d5671-7d7b-496a-bbdb-3eaa08be2e48 ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/44441#018d5671-7d7a-447e-a2d4-d765dc5ac918 ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/44441#018d5671-7d7b-4c3f-b1a8-5126d06818d0 ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/44441#018d5671-7d7c-40cf-8ab0-812dbfb4ac7c |
/dt |
204e1f5
to
ea86488
Compare
ea86488
to
6a75ebc
Compare
Previously we used a simple `std::vector` of futures to make waiting for partition leaders concurrent. Using a vector has a drawback when dealing with large number of topics and partitions since it may be required to allocate large contiguous chunk of memory for a future vector. In this particular case we may not use a fragmented vector or chunked fifo as the `when_all` uses a plain vector internally. To make sure no large chunk of memory is allocated to wait for the partition leaders changed the logic to use `seastar::max_concurrent_for_each`. Fixes: redpanda-data#15908 Signed-off-by: Michal Maslanka <michal@redpanda.com>
Sometimes it may happen that producer swarm is stopped after topic is recreated leading to a test failure. Added check restarting the producer if necessary Signed-off-by: Michal Maslanka <michal@redpanda.com>
6a75ebc
to
a18fdcd
Compare
/backport v23.3.x |
/backport v23.2.x |
Oops! Something went wrong. |
Oops! Something went wrong. |
/backport v23.3.x |
/backport v23.2.x |
Previously we used a simple
std::vector
of futures to makewaiting for partition leaders concurrent. Using a vector has a drawback
when dealing with large number of topics and partitions since it may be
required to allocate large contiguous chunk of memory for a future
vector. In this particular case we may not use a fragmented vector or
chunked fifo as the
when_all
uses a plain vector internally.To make sure no large chunk of memory is allocated to wait for the
partition leaders changed the logic to use
seastar::max_concurrent_for_each
.Fixes: #15908, #16270
Fixes: #16036
Backports Required
Release Notes