Emit per workload labels for existing per table vttablet metrics

[ejortegau]

Description

This adds the possibility to configure vttablet (via CLI flag) to also have a workload label for existing per table metrics (query counts, query times, query errors, query rows affected, query rows returned, query error counts). Workload can be any string that makes sense for the client application. For example, API endpoint name, controller, batch job name, application name or something else.

This is useful to be able to gain observability about how the query load is distributed across different workloads.

This is achieved with a new CLI flag, namely:

enable-per-workload-table-metrics: whether to enable or disable per workload metric collection - disabled by default to preserve the current behavior, thus making the new feature opt-in only. This flag is for vttablet.

The workload is obtained by parsing query directive comments of the form:

/*vt+ ... WORLOAD_NAME=<workload_name> ... */

For example, if vttablet is started with --enable-per-workload-table-metrics, and query is issued with a comment like

/*vt+ ... WORKLOAD_NAME=shop ... */

then metrics will look like

vttablet_query_counts{plan="Select",table="dual", workload="shop"} 15479

instead of

vttablet_query_counts{plan="Select",table="dual"} 15479

The workload name is also being added to the tracing span.

Related Issue(s)

#12490

Checklist

• "Backport to:" labels have been added if this change should be back-ported
• Tests were added or are not required
• Documentation was added or is not required

Deployment Notes

The new behavior with respect to metrics is disabled by default, and will only be enabled if the relevant CLI flag is passed when starting the vttablet process.

[vitess-bot]

Review Checklist

Hello reviewers! 👋 Please follow this checklist when reviewing this Pull Request.

General

• Ensure that the Pull Request has a descriptive title.
• If this is a change that users need to know about, please apply the release notes (needs details) label so that merging is blocked unless the summary release notes document is included.
• If a test is added or modified, there should be a documentation on top of the test to explain what the expected behavior is what the test does.

If a new flag is being introduced:

• Is it really necessary to add this flag?
• Flag names should be clear and intuitive (as far as possible)
• Help text should be descriptive.
• Flag names should use dashes (-) as word separators rather than underscores (_).

If a workflow is added or modified:

• Each item in Jobs should be named in order to mark it as required.
• If the workflow should be required, the maintainer team should be notified.

Bug fixes

• There should be at least one unit or end-to-end test.
• The Pull Request description should include a link to an issue that describes the bug.

Non-trivial changes

• There should be some code comments as to why things are implemented the way they are.

New/Existing features

• Should be documented, either by modifying the existing documentation or creating new documentation.
• New features should have a link to a feature request issue or an RFC that documents the use cases, corner cases and test cases.

Backward compatibility

• Protobuf changes should be wire-compatible.
• Changes to _vt tables and RPCs need to be backward compatible.
• vtctl command output order should be stable and awk-able.
• RPC changes should be compatible with vitess-operator
• If a flag is removed, then it should also be removed from VTop, if used there.

[deepthi]

@vitessio/query-serving this needs review from y'all..

[harshit-gangal]

I think we can avoid both of these flags. If the workload label is passed to vttablet that should be used otherwise ignored.

[deepthi]

I believe the idea here is that the label itself is customizable. If your query comment has workload: web-app, then the workload-label flag value would be workload. If you want to use xyzzy as the label, you can do that too. The other option is to hardcode the label as workload or some such pre-defined label. @ejortegau's approach is more flexible at the cost of adding another flag, so that's the trade-off.

[ejortegau]

Correct, I would like to have the label user-customizable as different folks/companies may already be sending some for of this with different labels; or may refer to this in different ways. I will, however, move this to the vtgate flags and obtain the workload information there, as suggested in the other comments.

[ejortegau]

I would like to keep the flag for enabling or disabling per workload metrics as this needs to be set up during statistic initialization to define whether there are 2 or 3 labels for the metrics involved - and this is before queries come in, which is when we could check whether the workload is passed to vttablet by vtgate or not. I have moved the other flag to vtgate instead, to allow the user to specify the key to look for in query directlives.

[frouioui]

LGTM. The docs should be change to document this new behavior and the new flag.

[ejortegau]

LGTM. The docs should be change to document this new behavior and the new flag.

Docs already include both behavior and flag in this PR.