add request ID injection to context to enable tracking requests across services

[erlan-z]

What this PR does:
This PR adds support for injecting a request ID into the request context to enable tracking requests across services.

• If the configured request_id header is present and non-empty, its value is used as the request ID.
• Otherwise, a new UUID is generated. For ruler queries, requestId is always generated.
• The request ID is stored in the context, accessible via requestmeta.RequestIdFromContext(r.Context()).

This ID is propagated through the service call chain and enables future tracking of downstream operations tied to a single originating request.

To enable request ID propagation, we followed the same mechanism previously used for targetHeaders propagation. As part of this change, we refactored the existing propagation logic to avoid duplication and to make it extensible for future request metadata.

For example, we may later want to add a request-origin field to the context to differentiate between ruler-initiated and ad-hoc queries at the storage layer. With the new shared request metadata structure, such additions can be made with minimal changes to the surrounding code.

Note: While there is no active usage of the request ID yet, the immediate goal is to support tracing storage-related resource usage—particularly identifying heavy or costly queries that stem from a single logical request.

Which issue(s) this PR fixes:
Fixes #

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]

Logs from testsing split queries got same request id

• query-frontend logs:

ts=2025-07-24T23:14:11.965255Z caller=handler.go:314 level=error header1=val1 header2=val2 x-cortex-request-id=request-id-sharded4 org_id=fake msg="query request received in QFE handler" requestId=request-id-sharded4

• querier logs:

ts=2025-07-24T23:14:11.971059Z caller=scheduler_processor.go:161 level=error header1=val1 header2=val2 x-cortex-request-id=request-id-sharded4 org_id=fake msg="query request received in Querier" requestId=request-id-sharded4
ts=2025-07-24T23:14:11.971067Z caller=scheduler_processor.go:161 level=error header1=val1 header2=val2 x-cortex-request-id=request-id-sharded4 org_id=fake msg="query request received in Querier" requestId=request-id-sharded4
ts=2025-07-24T23:14:11.97107Z caller=scheduler_processor.go:161 level=error header1=val1 header2=val2 x-cortex-request-id=request-id-sharded4 org_id=fake msg="query request received in Querier" requestId=request-id-sharded4

• ingester logs:

ts=2025-07-24T23:14:11.980543Z caller=ingester.go:2208 level=error msg="query request received in Ingester" requestId=request-id-sharded4
ts=2025-07-24T23:14:11.980543Z caller=ingester.go:2208 level=error msg="query request received in Ingester" requestId=request-id-sharded4
ts=2025-07-24T23:14:11.980543Z caller=ingester.go:2208 level=error msg="query request received in Ingester" requestId=request-id-sharded4

• store-gateway:

ts=2025-07-24T23:14:11.985171Z caller=gateway.go:414 level=error msg="query request received in SG" requestId=request-id-sharded4
ts=2025-07-24T23:14:11.985168Z caller=gateway.go:414 level=error msg="query request received in SG" requestId=request-id-sharded4
ts=2025-07-24T23:14:11.985173Z caller=gateway.go:414 level=error msg="query request received in SG" requestId=request-id-sharded4

• ruler logs:

ts=2025-07-24T23:28:39.885103Z caller=compat.go:193 level=error msg="query request id in ruler request" requestId=1d2f6e12-37b5-40fd-8a75-1021ff7ed114

• ingester logs:

ts=2025-07-24T23:28:39.887861Z caller=ingester.go:2208 level=error msg="query request received in Ingester" requestId=1d2f6e12-37b5-40fd-8a75-1021ff7ed114

[justinjung04]

Nice!

[justinjung04]

What's the goal here? Do you imagine a downstream component changing the header?

[erlan-z]

No specific case—just a precaution to avoid changing the context if the value already exists. This shouldn't normally happen, but in case it does, it's safer not to override it.

[justinjung04]

Thx for the logs from testing!

[danielblando]

The code lgtm, but it feels we are rewriting code from the log headers. The logs header is specific for logging, but one of the key parts was transferring ctx between services which seems to be duplicating here. I think we can reuse at least that logic

What if, we have a new config for the requestId header, but use the same middleware but adding the new header as targetHeader?
we can also refactor that util_log context function to a generic util.
would that also work? seems we would avoid some of this new changes

[danielblando]

nit: s/reqeusts/requests

[erlan-z]

The code lgtm, but it feels we are rewriting code from the log headers. The logs header is specific for logging, but one of the key parts was transferring ctx between services which seems to be duplicating here. I think we can reuse at least that logic

What if, we have a new config for the requestId header, but use the same middleware but adding the new header as targetHeader? we can also refactor that util_log context function to a generic util. would that also work? seems we would avoid some of this new changes

Thanks for feedback — changed it to use the same logic for transferring context between services as with targetHeaders. Also refactored the util_log context function to keep it log-specific, and moved the context propagation part into a separate util for request metadata. This keeps the concerns clearly separated and makes it easier to extend in the future.