define client each handler invocation

[ellenmuller]

What does this change?

Yesterday in the AWS office hour, we were told that the bedrock Client TCP connection is probably timing out after it is waiting for a response for bedrock for a long time. If true, this will exacerbate our backfill issues because it means that once we encounter the first hanging bedrock request, subsequent requests will always time out too because of the issue with the client.

Re-initialising the client each handler invocation won't fix the issues Bedrock server-side, but this might for now make our DLQ re-drive run more smoothly. It's worth a try!

See - https://aws.amazon.com/blogs/networking-and-content-delivery/implementing-long-running-tcp-connections-within-vpc-networking/

[github-actions]

Deploy build 14161 of media-service::grid::all to TEST

All deployment options

• Deploy build 14161 of media-service::grid::all to TEST
• Deploy parts of build 14161 to TEST by previewing it first
• What's on TEST right now?

---

From guardian/actions-riff-raff.

[joelochlann]

Nice! Thanks for actioning so quickly.

Only thing we might want to add is some timing logging around the client init, so we know how much it costs us in latency to re-init on every invocation. Could be useful for pressuring AWS to fix their dodgy SDK!

[gu-prout]

Seen on auth, usage, image-loader, metadata-editor, thrall, leases, cropper, collections, media-api, kahuna (merged by @ellenmuller 9 minutes and 46 seconds ago) Please check your changes!