backport (query: fix deadlock in endpointset (#4795))

[aymericDD]

Avoid RLock()ing twice as described here:
#4766 (comment)
(due to
https://stackoverflow.com/questions/30547916/goroutine-blocks-when-calling-rwmutex-rlock-twice-after-an-rwmutex-unlock/30549188).
Fix it by removing HasClients() and simply changing it with er.clients != nil.

Signed-off-by: Giedrius Statkevičius [email protected]

• I added CHANGELOG entry for this change.
• Change is not relevant to the end user.

Changes

Verification

[GiedriusS]

Thx, let's add the DCO 👍 cc @squat

[squat]

Sounds good to me. I can cut the next patch release for 0.23 and add this to the change log

[aymericDD]

@GiedriusS I cannot pass the DCO step because it is your commit.

[GiedriusS]

@GiedriusS I cannot pass the DCO step because it is your commit.

Ofc you can, just add your DCO on top with the suggested commands ;P

[aymericDD]

My bad 😇 My DCO is added to the commit

[squat]

Thanks @aymericDD (:
The docs check is failing because of this recent PR: #4896
It seems that when we build docs, we are using the directory structure of the current commit but then trying to access the docs from main, which is not guaranteed to work, and indeed fails in this case because the main branch has moved this doc. If we want to get this PR to build, we need to cherry-pick this commit, which seems silly. I would say we merge this and then consider making a PR to the release-0.23 branch to fix CI.

WDYT @GiedriusS?

[bwplotka]

Makes sense, thanks!

[aymericDD]

@squat do you know when the fix release will be published? It is very painful for us to maintain Thanos in production. Our queries are stuck or restart after a long time and we have high latency....

[squat]

Today 👍

[aymericDD]

Thanks @squat 👍

[squat]

Here's the new PR to cut 0.23.2: #4936