feat(http2): configure header list size for tcp/uds

[aaronArinder]

---

Checklist

Complete the checklist (and note appropriate exceptions) before the PR is marked ready-for-review.

• PR description explains the motivation for the change and relevant context for reviewing
• PR description links appropriate GitHub/Jira tickets (creating when necessary)
• Changeset is included for user-facing changes
• Changes are compatible¹
• Documentation² completed
• Performance impact assessed and acceptable
• Metrics and logs are added³ and documented
• Tests added and passing⁴
  • Unit tests
  • Integration tests
  • Manual tests, as necessary

Exceptions

Note any exceptions here

Notes

Footnotes

1. It may be appropriate to bring upcoming changes to the attention of other (impacted) groups. Please endeavour to do this before seeking PR approval. The mechanism for doing this will vary considerably, so use your judgement as to how and when to do this. ↩

2. Configuration is an important part of many changes. Where applicable please try to document configuration examples. ↩

3. A lot of (if not most) features benefit from built-in observability and debug-level logs. Please read this guidance on metrics best-practices. ↩

4. Tick whichever testing boxes are applicable. If you are adding Manual Tests, please document the manual testing (extensively) in the Exceptions. ↩

[apollo-librarian]

✅ Docs preview ready

The preview is ready to be viewed. View the preview

File Changes

0 new, 1 changed, 0 removed

* graphos/routing/(latest)/performance/caching/response-caching/customization.mdx

Build ID: dacd44e79706b137ce76a4ad
Build Logs: View logs

URL: https://www.apollographql.com/docs/deploy-preview/dacd44e79706b137ce76a4ad

[bnjjj]

Could it happen that the header size would be bigger than u32::MAX ? If so I would prefer using u32::try_from() instead of as cast. WDYT ?

[aaronArinder]

technically yes, but practically I don't think so; there'd have to be ~4.3gb of headers for u32 (and almost 20k petabytes for usize), which would probably choke the router somewhere earlier; but, I've added a try_from() and a warning log just in case (because who knows what folks get up to)

[bnjjj]

Same here with the cast

[yliu-intuit]

Hi @aaronArinder , the fix didn't work with our TCP connections. We continue to receive 431 when sending 10KB single header despite setting the config of http2_max_headers_list_bytes: "48KiB".

The issue is fixed when specifically check for HTTP2 connection and apply http2_only() on the builder.

This is caused by using the auto-detection builder in this PR. It happens as follows as a possible explanation.

1. State Machine Initialization: When an HTTP/2 connection object is created, it must start with some rules for how much memory it allows the peer to use.

2. The Default: The h2 crate defaults the SETTINGS_MAX_HEADER_LIST_SIZE to 16,384 bytes (16KB).

3. The Update: If you configure a larger size (e.g., 800KB), the server must send a SETTINGS frame to the client to say "I actually accept 800KB."

4. The Gap: In "Auto" mode, the server buffers the incoming bytes to detect the protocol. Once it detects "Oh, this is HTTP/2," it creates the H2 connection object. Critically, if it parses the incoming data before it has time to apply your configuration overrides and process the SETTINGS update, it uses the Initial Default (16KB).

I have a fix that force using http2_only() and it solves the problem. you can check here. https://github.com/yliu-intuit/router/pull/1/files

However, I am still looking into why #8631 works. #8631 is using the same logic as in this PR.