Increase remote recovery thread pool size #10750
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
andrross
requested review from
abbashus,
anasalkouz,
Bukhtawar,
CEHENKLE,
dblock,
dbwiddis,
dreamer-89,
gbbafna,
kartg,
kotwanikunal,
mch2,
msfroh,
nknize,
owaiskazi19,
peternied,
reta,
Rishikesh1159,
ryanbogan,
sachinpkale,
saratvemulapalli,
setiah,
shwetathareja,
sohami and
VachaShah
as code owners
andrross
force-pushed
the
update-remote-recovery-thread-pool-size
branch
from
dc8a839 to
088dbf7
Compare
Gradle Check (Jenkins) Run Completed with:
|
mch2
approved these changes
Oct 19, 2023
kotwanikunal
approved these changes
Oct 19, 2023
Compatibility status:Checks if related components are compatible with change 5499b8b Incompatible componentsSkipped componentsCompatible componentsCompatible components: [https://github.com/opensearch-project/security-analytics.git, https://github.com/opensearch-project/custom-codecs.git, https://github.com/opensearch-project/security.git, https://github.com/opensearch-project/index-management.git, https://github.com/opensearch-project/opensearch-oci-object-storage.git, https://github.com/opensearch-project/geospatial.git, https://github.com/opensearch-project/job-scheduler.git, https://github.com/opensearch-project/sql.git, https://github.com/opensearch-project/notifications.git, https://github.com/opensearch-project/observability.git, https://github.com/opensearch-project/k-nn.git, https://github.com/opensearch-project/neural-search.git, https://github.com/opensearch-project/cross-cluster-replication.git, https://github.com/opensearch-project/alerting.git, https://github.com/opensearch-project/performance-analyzer.git, https://github.com/opensearch-project/anomaly-detection.git, https://github.com/opensearch-project/performance-analyzer-rca.git, https://github.com/opensearch-project/asynchronous-search.git, https://github.com/opensearch-project/ml-commons.git, https://github.com/opensearch-project/common-utils.git, https://github.com/opensearch-project/reporting.git] |
Gradle Check (Jenkins) Run Completed with:
|
reta
reviewed
Oct 19, 2023
server/src/main/java/org/opensearch/indices/recovery/RecoverySettings.java
Show resolved
Hide resolved
andrross
force-pushed
the
update-remote-recovery-thread-pool-size
branch
from
088dbf7 to
a60da9b
Compare
Gradle Check (Jenkins) Run Completed with:
|
andrross
force-pushed
the
update-remote-recovery-thread-pool-size
branch
from
a60da9b to
5155700
Compare
The remote recovery thread pool does blocking I/O when downloading files, so the "half processor count max 10" was definitely too small. This can be shown by triggering recoveries on a node that is also doing segment replication, and the replication lag will increase due to contention on that thread pool. Some amount of contention is inevitable, but the change here to increase the download thread pool, and also limit the concurrent usage of that thread pool by any single recovery/replication to 25% of the threads does help. Long term, we can improve this even further by moving to fully async I/O to avoid blocking threads in the application on draining InputStreams. Signed-off-by: Andrew Ross <[email protected]>
andrross
force-pushed
the
update-remote-recovery-thread-pool-size
branch
from
5155700 to
5499b8b
Compare
Gradle Check (Jenkins) Run Completed with:
|
Gradle Check (Jenkins) Run Completed with:
|
opensearch-trigger-bot bot
pushed a commit
that referenced
this pull request
Oct 20, 2023
The remote recovery thread pool does blocking I/O when downloading files, so the "half processor count max 10" was definitely too small. This can be shown by triggering recoveries on a node that is also doing segment replication, and the replication lag will increase due to contention on that thread pool. Some amount of contention is inevitable, but the change here to increase the download thread pool, and also limit the concurrent usage of that thread pool by any single recovery/replication to 25% of the threads does help. Long term, we can improve this even further by moving to fully async I/O to avoid blocking threads in the application on draining InputStreams. Signed-off-by: Andrew Ross <[email protected]> (cherry picked from commit 1e28738) Signed-off-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
mch2
pushed a commit
that referenced
this pull request
Oct 21, 2023
The remote recovery thread pool does blocking I/O when downloading files, so the "half processor count max 10" was definitely too small. This can be shown by triggering recoveries on a node that is also doing segment replication, and the replication lag will increase due to contention on that thread pool. Some amount of contention is inevitable, but the change here to increase the download thread pool, and also limit the concurrent usage of that thread pool by any single recovery/replication to 25% of the threads does help. Long term, we can improve this even further by moving to fully async I/O to avoid blocking threads in the application on draining InputStreams. (cherry picked from commit 1e28738) Signed-off-by: Andrew Ross <[email protected]> Signed-off-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com> Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
austintlee
pushed a commit
to austintlee/OpenSearch
that referenced
this pull request
Oct 23, 2023
The remote recovery thread pool does blocking I/O when downloading files, so the "half processor count max 10" was definitely too small. This can be shown by triggering recoveries on a node that is also doing segment replication, and the replication lag will increase due to contention on that thread pool. Some amount of contention is inevitable, but the change here to increase the download thread pool, and also limit the concurrent usage of that thread pool by any single recovery/replication to 25% of the threads does help. Long term, we can improve this even further by moving to fully async I/O to avoid blocking threads in the application on draining InputStreams. Signed-off-by: Andrew Ross <[email protected]>
shiv0408
pushed a commit
to Gaurav614/OpenSearch
that referenced
this pull request
Apr 25, 2024
The remote recovery thread pool does blocking I/O when downloading files, so the "half processor count max 10" was definitely too small. This can be shown by triggering recoveries on a node that is also doing segment replication, and the replication lag will increase due to contention on that thread pool. Some amount of contention is inevitable, but the change here to increase the download thread pool, and also limit the concurrent usage of that thread pool by any single recovery/replication to 25% of the threads does help. Long term, we can improve this even further by moving to fully async I/O to avoid blocking threads in the application on draining InputStreams. Signed-off-by: Andrew Ross <[email protected]> Signed-off-by: Shivansh Arora <[email protected]>
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.
The remote recovery thread pool does blocking I/O when downloading files, so the "half processor count max 10" was definitely too small. This can be shown by triggering recoveries on a node that is also doing segment replication, and the replication lag will increase due to contention on that thread pool. Some amount of contention is inevitable, but the change here to increase the download thread pool, and also limit the concurrent usage of that thread pool by any single recovery/replication to 25% of the threads does help.
Long term, we can improve this even further by moving to fully async I/O to avoid blocking threads in the application on draining InputStreams.
Check List
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.