Consul job rename/relabel not working with 2.5.0

[raypettersen]

Proposal

Use case. Why is this important?
Grouping of exporters when using consul is broken for job rename.

Bug Report

What did you do?
Upgraded to latest 2.5 version. All exporters discovered with consul ended up with a single job name instead of rename for each exporter which was working flawlessly with 1.8.0

What did you expect to see?
Continued renaming of job name, so all discovered exporters are sorted correctly in the prometheus view. Downgrading to 2.4.3 resolved the issue.

• Prometheus version:

2.5.0

[raypettersen]

        regex: '(.*)'
        target_label: 'job'
        replacement: '$1'

[hoffie]

Do you see any actual problems (i.e. unexpected changes in the database) or could this be just explained by the announced, changed UI grouping in 2.5.0?

[raypettersen]

It's just the gui as far as I can see. Likely related to the announcement. What's the fix for getting proper grouping when using consul/sd? Right now everything is grouped under "consul". I'd like to get the old behavior which lets me get exporters grouped by name.

[simonpasquier]

Indeed the UI layout for targets has changed in 2.5.0. To be honest, it wasn't really intentional but positively welcomed by users that noticed the new layout. On the other hand, I see how the change can be annoying for you. What if targets inside a scrape configuration were grouped by job?

[MustaphaB1]

Also having this same issue.. will this be fixed or is downgrading to 2.4.3 the only option.

[simonpasquier]

I've sent #4898 that will at least sort targets by job then instance labels.

[zbindenren]

Also having this same issue.. will this be fixed or is downgrading to 2.4.3 the only option.

Downgrading to 2.4 is the only option to get the same behavior.

[raypettersen]

Problem persists with 2.5, with "standard" consul config.

[zbindenren]

@raypettersen the problem also persists with 2.6. And there are no plans to go back to the
<= 2.4 behavior. See #4898 for more information.

[raypettersen]

Yes, we're staying on 2.4.3 until we find a workaround. I thought #4898 would resolve the problem, but it didn't.