Resort store response set on internal label dedup #6529
Conversation
…s own minio and StoreGW. The main motivation for doing this is that each testcase needs to allow StoreGW to advertise a different set of labels as external labels, and that no two cases interfere with each other. Signed-off-by: Saswata Mukherjee <[email protected]>
When deduplicating on labels which are stored internally in TSDB, the store response set needs to be resorted after replica labels are removed. This commit detects when internal labels are being used for dedup, by checking for their presence in the Store client's external labels. Signed-off-by: Saswata Mukherjee <[email protected]>
Sonatype Lift is retiringSonatype Lift will be retiring on Sep 12, 2023, with its analysis stopping on Aug 12, 2023. We understand that this news may come as a disappointment, and Sonatype is committed to helping you transition off it seamlessly. If you’d like to retain your data, please export your issues from the web console. |
| @@ -618,6 +618,25 @@ func newAsyncRespSet( | |||
| } | |||
| } | |||
|
|
|||
| // hasInternalReplicaLabels returns true if any replica label in the series request is not an | |||
| // external label for the given Client. | |||
| func hasInternalReplicaLabels(st Client, req *storepb.SeriesRequest) bool { | |||
There was a problem hiding this comment.
This will always lead to re-sorting if stores have different external labels right? Like prometheus_replica on a sidecar and rule_replica on a ruler.
Is this relevant for the test? Is there any interference? Spinning up all these Store GWs, MinIOs, and uploading on each test case will take much more time relative to the way this is done right now. I was intentionally trying to save time, especially as I noticed using a single Store GW I was able to reproduce all scenarios and didn't identify any bad interference. I think we might potentially want to keep both scenarios. |
|
Yup it was interfering, but I think that's due to the nature of this fix. We match replica labels to store external labels exactly, and fall back otherwise. I modified it to per test case, so that the store gw exposes the exact replica labels needed for the test. But yeah this can probably be reverted in the bloom filter approach. |
|
@douglascamata reverted this test change in #6317. All these tests now pass with correct values there 🙂 |
Changes
This PR does a couple of things,
TestQueryStoreDedupto start a new StoreGW and minio instance on every testcase. The main motivation for doing this is that each testcase needs to allow StoreGW to advertise a different set of labels as external labels and that no two cases interfere with each other.Aims to fix #6257. Takes after @fpetkovski's earlier work in #6317 (but inverts the approach).
Verification
Modified regression test cases for Store, Sidecar and Receive dedup and tested locally.