chore(server): Set keep-alive and headers timeouts on server start

[ikoenigsknecht]

No description provided.

[goldcaddy77]

Thanks for the PR! I already expose the httpServer on Server object, so you could do the following:

import { Server } from 'warthog';

const app = new Server();
app.start();
app.httpServer.keepAliveTimeout = keepAliveTimeout;
app.httpServer.headersTimeout = headersTimeout;

Do you know if most folks change these values from the defaults? If so, your change probably makes sense.

[ikoenigsknecht]

That's what we are currently doing and setting those values stopped our random 502 issue. It's also what we do elsewhere at Indigo. I'm not sure if this is what everyone does but 30s+ has been standard everywhere I've worked. My goal was to make it simpler and easier for others to avoid the same performance problems that we encountered.

[goldcaddy77]

I think since you changed this to not be an offset, we'd need to set this to a value larger than WARTHOG_KEEP_ALIVE_TIMEOUT_MS, right?

In looking at these pages:

• https://shuheikagawa.com/blog/2019/04/25/keep-alive-timeout/
• https://docs.aws.amazon.com/elasticloadbalancing/latest/application/application-load-balancers.html#connection-idle-timeout
• https://nodejs.org/api/http.html#http_server_headerstimeout

It seems both of these values need to be > 60. Should we do something like 61000 and 62000?

[ikoenigsknecht]

It's more about headersTimeout > keepAliveTimeout. 30s is a good keep alive (though if you want it to be 60s by default I can certainly make that change), but the key part of this is to ensure that the headers time out is greater than the keep alive.

[ikoenigsknecht]

I pushed up the change to make headers timeout 60s, that was an oversight in my last commit.

[goldcaddy77]

Looks great, thanks 👍

[goldcaddy77]

🎉 This PR is included in version 2.7.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀