Open hanem100k opened 2 months ago
Note: https://github.com/thanos-io/thanos/blob/883fade9bd75fe595b6e947a33c59e27fca1abda/pkg/query/remote_engine.go#L116 likely the MinT() computation is off for usecases where we have [sidecar,receiver,..]+storagegw endpoints on same remote engine.
@hanem100k can you try attaching a separate querier to store gateways?
I do have a Top level [QFE] -> [Q] -> Leaf [Q] -> [receivers, stores]
There is a single Query in the top layer and there are N leafs.
If I go into the leaf Querier directly, it returns ok from stores as well as receivers.
@fpetkovski
can you try attaching a separate querier to store gateways?
Do you mean that I should retry the distributed mode by having a Querier for stores only and one for receivers only?
Actually we are going to fix this configuration with @MichaHoffmann soon, so let's wait for the new patch release.
@fpetkovski Is this issue already fixed?
Query, store gateway
Thanos, Prometheus and Golang version used: Thanos version 0.26.1
Object Storage Provider: Azure blob store
What happened: Having a set of receivers and sharded store gateways. Whenever query distributed mode is enabled, data will only be returned from receivers but not stores.
Removing
query.mode=distributed
from arguments solves the issue.What you expected to happen:
How to reproduce it (as minimally and precisely as possible): It is reproducible 100% when you query for a time range that exceeds your receivers.
Metrics with distribution enabled:
Metrics without distribution enabled: