Kafka sink to open fewer threads #431
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
istreeter
force-pushed
the
kafka-sink-fewer-threads
branch
from
cd45a9d to
e585e8e
Compare
istreeter
force-pushed
the
kafka-sink-fewer-threads
branch
from
e585e8e to
6abd80b
Compare
peel
approved these changes
Nov 13, 2024
AlexBenny
pushed a commit
that referenced
this pull request
Jan 7, 2025
In #431 we improved performance of the Kafka sink by calling `producer.send()` on a compute thread not on a blocking thread. This is a great improvement if it is always true that `producer.send()` never blocks. But `producer.send()` does in fact block if the producer needs to re-fetch topic metadata. This happens every 5 minutes by default, and is configured by the kafka setting `metadata.max.age.ms`. If `producer.send()` blocks on a compute thread, then it can potentially cause thread starvation, and negatively impact the collector's responsiveness to requests. This PR changes to executing `send()` on a dedicated thread. From the point of view of the collector, it is ok if `send()` is blocking on the dedicated thread. It is better than running it inside `Sync[F].blocking` (which is what we did before) because that tended to create a huge number of threads under some circumstances. In theory this shouldn't negatively affect performance much, even though it's a single thread. Because most of the time `send()` does not block; and when it does block (i.e. once per 5 minutes) then it is ok for the other events to get enqueued as a backlog of Callables on the thread. Kafka sink use a dedicated thread for potentially blocking send: Amendment 1
AlexBenny
pushed a commit
that referenced
this pull request
Jan 7, 2025
In #431 we improved performance of the Kafka sink by calling `producer.send()` on a compute thread not on a blocking thread. This is a great improvement if it is always true that `producer.send()` never blocks. But `producer.send()` does in fact block if the producer needs to re-fetch topic metadata. This happens every 5 minutes by default, and is configured by the kafka setting `metadata.max.age.ms`. If `producer.send()` blocks on a compute thread, then it can potentially cause thread starvation, and negatively impact the collector's responsiveness to requests. This PR changes to executing `send()` on a dedicated thread. From the point of view of the collector, it is ok if `send()` is blocking on the dedicated thread. It is better than running it inside `Sync[F].blocking` (which is what we did before) because that tended to create a huge number of threads under some circumstances. In theory this shouldn't negatively affect performance much, even though it's a single thread. Because most of the time `send()` does not block; and when it does block (i.e. once per 5 minutes) then it is ok for the other events to get enqueued as a backlog of Callables on the thread. Kafka sink use a dedicated thread for potentially blocking send: Amendment 1
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.