Label Values composite index

Question

Label Values composite index

gouthamve opened this issue 8 years ago · 7 comments

I see from this doc that we are going to have composite indexes. But it is still un-implemented.

@fabxc I am assuming this is a TODO item and you have not dropped the idea of a composite index.

Answer 1 · 2017-03-29T20:41:22.000Z

Yes, this is a to-do that should be possible to add seemlessly at a later point. The label value index interface shows support for it already. It is useful for advanced autocompletion in Prometheus. For example, if we want to suggest values for a certain label, it makes sense to only suggest those that exist for the given metric name (__name__ label). The composite index provides the required lookup.

…

On Wed, Mar 29, 2017, 9:52 PM Goutham Veeramachaneni < ***@***.***> wrote: I see from this <https://docs.google.com/document/d/13uK6HoEXzX67DuWsJ2PKo8zjpqlgpQ0SIHz2nvMa7f0/edit#heading=h.zzd5egnpdp5> doc that we are going to have composite indexes. But it is still un-implemented. @fabxc <https://github.com/fabxc> I am assuming this is a TODO item and you have not dropped the idea of a composite index. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#26>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEuA8iGyoy3yj2qwMCQVLMlMTxytn2b9ks5rqraOgaJpZM4MteST> .

Answer 2 · 2018-08-31T10:23:38.000Z

@fabxc , @gouthamve
having red the issue description briefly this is still in need to be implemented right?
just so I know how to label the issue.

Answer 3 · 2018-08-31T10:29:14.000Z

@krasi-georgiev I was about to implement this in my GSoC. But as there were not many use cases with this, it was decided that we don't implement it for now.

Answer 4 · 2018-08-31T10:33:46.000Z

yeah that is what I wanted to know.
Autocompletion sounds very useful so really surprising why it didn't get more interest.

Answer 5 · 2018-08-31T11:20:28.000Z

Well, there was no use-case for this. Current APIs also allow for autocompletion but this was supposed to make it more efficient.

This means hardcoding what to autocomplete against and the current implementations were fast enough to not require this optimisation.

Answer 6 · 2018-08-31T11:25:21.000Z

so in summary

what we have works well enough
there isn't any immediate demand for this

Seems we close this for now and revisit if needed?

Answer 7 · 2018-08-31T11:30:55.000Z

Yes. Also @codesome has a branch somewhere that does this, I think (: