Loki: introduce an exponential backoff on failed query retries. #10959
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
Signed-off-by: Edward Welch <edward.welch@grafana.com>
Signed-off-by: Edward Welch <edward.welch@grafana.com>
owen-d
approved these changes
Oct 18, 2023
dannykopping
approved these changes
Oct 18, 2023
| @@ -70,7 +86,8 @@ func (r retry) Do(ctx context.Context, req Request) (Response, error) { | |||
| httpResp, ok := httpgrpc.HTTPResponseFromError(err) | |||
| if !ok || httpResp.Code/100 == 5 { | |||
| lastErr = err | |||
| level.Error(util_log.WithContext(ctx, r.log)).Log("msg", "error processing request", "try", tries, "query", req.GetQuery(), "err", err) | |||
| level.Error(util_log.WithContext(ctx, r.log)).Log("msg", "error processing request", "try", tries, "query", req.GetQuery(), "retry_in", bk.NextDelay(), "err", err) | |||
There was a problem hiding this comment.
NextDelay() 👍 nice touch
slim-bean
added a commit
that referenced
this pull request
Oct 25, 2023
**What this PR does / why we need it**: Currently any subquery which fails with a 5xx status code is retried by the query frontend immediately, and by default will make a total of 5 attempts to execute the subquery, ultimately failing the query if you exhaust those attempts. We notice there are times when subqueries can fail extremely quickly, mostly around queriers handling various GRPC related issues between components like ingesters or index gateways. In these cases a query will exhaust all the retries in less than a second and fail back to the user. Also in cases where a query is causing a lot of memory pressure on a pool of queriers leading to OOM crashes of queriers, the lack of a backoff on retries can make this behavior quite aggressive and potentially cause disruption if you don't have enough queriers available. This PR introduces a simple exponential backoff on retries of failed queries to allow for the downstream GRPC issues to have time to sort out as well as causing queries causing a lot of memory pressure and OOMing to be slowed down and minimize their impact slightly. **Which issue(s) this PR fixes**: Fixes #<issue number> **Special notes for your reviewer**: **Checklist** - [ ] Reviewed the [`CONTRIBUTING.md`](https://github.com/grafana/loki/blob/main/CONTRIBUTING.md) guide (**required**) - [ ] Documentation added - [ ] Tests updated - [x] `CHANGELOG.md` updated - [ ] If the change is worth mentioning in the release notes, add `add-to-release-notes` label - [ ] Changes that require user attention or interaction to upgrade are documented in `docs/sources/setup/upgrade/_index.md` - [ ] For Helm chart changes bump the Helm chart version in `production/helm/loki/Chart.yaml` and update `production/helm/loki/CHANGELOG.md` and `production/helm/loki/README.md`. [Example PR](d10549e) --------- Signed-off-by: Edward Welch <edward.welch@grafana.com> (cherry picked from commit d66dd30)
slim-bean
added a commit
that referenced
this pull request
Oct 25, 2023
**What this PR does / why we need it**: Currently any subquery which fails with a 5xx status code is retried by the query frontend immediately, and by default will make a total of 5 attempts to execute the subquery, ultimately failing the query if you exhaust those attempts. We notice there are times when subqueries can fail extremely quickly, mostly around queriers handling various GRPC related issues between components like ingesters or index gateways. In these cases a query will exhaust all the retries in less than a second and fail back to the user. Also in cases where a query is causing a lot of memory pressure on a pool of queriers leading to OOM crashes of queriers, the lack of a backoff on retries can make this behavior quite aggressive and potentially cause disruption if you don't have enough queriers available. This PR introduces a simple exponential backoff on retries of failed queries to allow for the downstream GRPC issues to have time to sort out as well as causing queries causing a lot of memory pressure and OOMing to be slowed down and minimize their impact slightly. **Which issue(s) this PR fixes**: Fixes #<issue number> **Special notes for your reviewer**: **Checklist** - [ ] Reviewed the [`CONTRIBUTING.md`](https://github.com/grafana/loki/blob/main/CONTRIBUTING.md) guide (**required**) - [ ] Documentation added - [ ] Tests updated - [x] `CHANGELOG.md` updated - [ ] If the change is worth mentioning in the release notes, add `add-to-release-notes` label - [ ] Changes that require user attention or interaction to upgrade are documented in `docs/sources/setup/upgrade/_index.md` - [ ] For Helm chart changes bump the Helm chart version in `production/helm/loki/Chart.yaml` and update `production/helm/loki/CHANGELOG.md` and `production/helm/loki/README.md`. [Example PR](d10549e) --------- Signed-off-by: Edward Welch <edward.welch@grafana.com> (cherry picked from commit d66dd30)
rhnasc
pushed a commit
to inloco/loki
that referenced
this pull request
Apr 12, 2024
…ana#10959) **What this PR does / why we need it**: Currently any subquery which fails with a 5xx status code is retried by the query frontend immediately, and by default will make a total of 5 attempts to execute the subquery, ultimately failing the query if you exhaust those attempts. We notice there are times when subqueries can fail extremely quickly, mostly around queriers handling various GRPC related issues between components like ingesters or index gateways. In these cases a query will exhaust all the retries in less than a second and fail back to the user. Also in cases where a query is causing a lot of memory pressure on a pool of queriers leading to OOM crashes of queriers, the lack of a backoff on retries can make this behavior quite aggressive and potentially cause disruption if you don't have enough queriers available. This PR introduces a simple exponential backoff on retries of failed queries to allow for the downstream GRPC issues to have time to sort out as well as causing queries causing a lot of memory pressure and OOMing to be slowed down and minimize their impact slightly. **Which issue(s) this PR fixes**: Fixes #<issue number> **Special notes for your reviewer**: **Checklist** - [ ] Reviewed the [`CONTRIBUTING.md`](https://github.com/grafana/loki/blob/main/CONTRIBUTING.md) guide (**required**) - [ ] Documentation added - [ ] Tests updated - [x] `CHANGELOG.md` updated - [ ] If the change is worth mentioning in the release notes, add `add-to-release-notes` label - [ ] Changes that require user attention or interaction to upgrade are documented in `docs/sources/setup/upgrade/_index.md` - [ ] For Helm chart changes bump the Helm chart version in `production/helm/loki/Chart.yaml` and update `production/helm/loki/CHANGELOG.md` and `production/helm/loki/README.md`. [Example PR](grafana@d10549e) --------- Signed-off-by: Edward Welch <edward.welch@grafana.com>
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.
What this PR does / why we need it:
Currently any subquery which fails with a 5xx status code is retried by the query frontend immediately, and by default will make a total of 5 attempts to execute the subquery, ultimately failing the query if you exhaust those attempts.
We notice there are times when subqueries can fail extremely quickly, mostly around queriers handling various GRPC related issues between components like ingesters or index gateways.
In these cases a query will exhaust all the retries in less than a second and fail back to the user.
Also in cases where a query is causing a lot of memory pressure on a pool of queriers leading to OOM crashes of queriers, the lack of a backoff on retries can make this behavior quite aggressive and potentially cause disruption if you don't have enough queriers available.
This PR introduces a simple exponential backoff on retries of failed queries to allow for the downstream GRPC issues to have time to sort out as well as causing queries causing a lot of memory pressure and OOMing to be slowed down and minimize their impact slightly.
Which issue(s) this PR fixes:
Fixes #
Special notes for your reviewer:
Checklist
CONTRIBUTING.mdguide (required)CHANGELOG.mdupdatedadd-to-release-noteslabeldocs/sources/setup/upgrade/_index.mdproduction/helm/loki/Chart.yamland updateproduction/helm/loki/CHANGELOG.mdandproduction/helm/loki/README.md. Example PR