Extend the list of status codes retried in client

[losipiuk]

This commit extends list of codes on which client will retry to:

• HTTP_BAD_GATEWAY (502)
• HTTP_UNAVAILABLE (503)
• HTTP_GATEWAY_TIMEOUT(504)

Release notes

(x) This is not user-visible or docs only and no release notes are required.
( ) Release notes are required, please propose a release note for me.
( ) Release notes are required, with the following suggested text:

[losipiuk]

updated - there was a bug there actully (== instead of !=)

[electrum]

I don't think we should retry arbitrary 5xx code. Not all are appropriate, for example

• 501 Not Implemented
• 505 HTTP Version Not Supported

We could retry on these:

• 502 Bad Gateway
• 503 Service Unavailable
• 504 Gateway Timeout

We also need to update docs/src/main/sphinx/develop/client-protocol.rst to match this change in the protocol specification.

[losipiuk]

Thanks for review @electrum, @findepi . Updated

[electrum]

I think we need to update the check in HttpTokenPoller as well, which will permanently fail the request if it is not 200 or 503.

[electrum]

Trino actually doesn't return 503. That was the intended code for gateways, etc., but in practice, we need to support 502, 503, 504, so we can just mention all three here. In other words, don't differentiate between them here, just list all three.

[losipiuk]

reworded docuementation.

[losipiuk]

I think we need to update the check in HttpTokenPoller

Good point. Updated that one too.

[losipiuk]

@electrum should be good to go now IMO

[sajjoseph]

Thanks for fixing this. I had this patch in our local version for a while as we had issues with trino-client code and the proxy layer before trino-server. A similar fix was suggested for trino-python-client as well.