thanos-query-frontend: Enable Thanos Query Stats Propagation & cache response headers #10
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
milesbxf
force-pushed
the
return-stats-response
branch
2 times, most recently
from
a5a0575 to
11d09c7
Compare
milesbxf
changed the title
milesbxf
force-pushed
the
return-stats-response
branch
9 times, most recently
from
acee5c0 to
de9c1b0
Compare
Thanos Query gives us some really nice detailed internal stats - but thanos-query-frontend annoyingly blats them by trying to parse the response as a Prometheus response, which has different fields and a different structure. In addition, thanos-query-frontend doesn't even pass the &stats=all parameter through if you try to request it. This change fixes this by propagating the stats request parameter to the downstream, and decoding/encoding the Thanos Query stats structure properly. As an extra, we also set a response header if we get a cache hit so that upstreams can use this. Signed-off-by: milesbryant <[email protected]>
milesbxf
force-pushed
the
return-stats-response
branch
from
de9c1b0 to
64eb68e
Compare
milesbxf
commented
Nov 18, 2024
| @@ -145,8 +145,6 @@ func registerQueryFrontend(app *extkingpin.App) { | |||
|
|
|||
| cmd.Flag("query-frontend.log-queries-longer-than", "Log queries that are slower than the specified duration. "+ | |||
| "Set to 0 to disable. Set to < 0 to enable on all queries.").Default("0").DurationVar(&cfg.CortexHandlerConfig.LogQueriesLongerThan) | |||
| cmd.Flag("query-frontend.query-stats-enabled", "True to enable query statistics tracking. "+ | |||
There was a problem hiding this comment.
I've removed this now that we return query stats directly in the response
| @@ -84,12 +84,22 @@ message Matrix { | |||
| } | |||
|
|
|||
| message PrometheusResponseStats { | |||
| PrometheusResponseSamplesStats samples = 1 [(gogoproto.jsontag) = "samples"]; | |||
| } | |||
| message Timings { | |||
There was a problem hiding this comment.
This is the difference between Thanos stats and Prometheus stats responses
steg
reviewed
Nov 18, 2024
| for _, existing := range headers { | ||
| if existing.Name == newHeader.Name { | ||
| // if headers match, overwrite with the new header | ||
| existing = newHeader |
There was a problem hiding this comment.
Should this be *existing = *newHeader ?
Also, below we sort the responses by response time and then the first non-null explanation is taken rather than the last one. Is consistency between these important?
| func (resp *PrometheusInstantQueryResponse) AddHeader(key, value string) { | ||
| resp.Headers = append(resp.Headers, &PrometheusResponseHeader{Name: key, Values: []string{value}}) | ||
| } | ||
|
|
There was a problem hiding this comment.
I think we'll end up with duplicate headers if the same header key is added twice - should this check for the header already existing and either replace it or append to its values?
| @@ -217,9 +227,25 @@ func (prometheusCodec) MergeResponse(_ Request, responses ...Response) (Response | |||
| // we need to pass on all the headers for results cache gen numbers. | |||
| var resultsCacheGenNumberHeaderValues []string | |||
|
|
|||
| var headers []*PrometheusResponseHeader | |||
| // merge headers | |||
| for _, res := range responses { | |||
| promResponses = append(promResponses, res.(*PrometheusResponse)) | |||
| resultsCacheGenNumberHeaderValues = append(resultsCacheGenNumberHeaderValues, getHeaderValuesWithName(res, ResultsCacheGenNumberHeaderName)...) | |||
There was a problem hiding this comment.
I doubt this matters given what was happening before, but this could be out of sync with the actual headers we return in the response due to the merging below.
| // Copy Prometheus headers into http response | ||
| for _, h := range a.Headers { | ||
| for _, v := range h.Values { | ||
| if strings.HasPrefix(h.Name, "X-") { |
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.
Thanos Query gives us some really nice detailed internal stats - but thanos-query-frontend annoyingly blats them by trying to parse the response as a Prometheus response, which has different fields and a different structure.
In addition, thanos-query-frontend doesn't even pass the &stats=all parameter through if you try to request it.
This change fixes this by propagating the stats request parameter to the downstream, and decoding/encoding the Thanos Query stats structure properly.
As an extra, we also set a response header if we get a cache hit so that upstreams can use this.