Resource should retry on ECONNRESET

[xoebus]

It looks like we used to retry on this error but commit 889a220 accidentally broke this as ECONNRESET is not temporary or a timeout.

I'm not 100% sure that the error will bubble up to the client as syscall.ECONNRESET but should we retry more aggressively on all error types anyway?

[vito]

Can we just use https://github.com/concourse/retryhttp?

[xoebus]

From README:

Retries on network errors, does not retry if request body was already read from (e.g. streaming request)

Would this cause it to not retry the above because it had already downloaded ~400MB?

[xoebus]

Maybe this is a retry loop outside a single logical HTTP request?

[zachgersh]

Are you saying just start the whole request over again if it fails? Zach

…

On Jul 31, 2019, 5:23 PM -0400, Christopher Brown ***@***.***>, wrote: Maybe this is a retry loop outside a single logical HTTP request? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

[xoebus]

Yep

[xoebus]

Not sure if this should be encapsulated inside the client or not but the lines below could be wrapped in a fixed-retry loop?

bosh-io-stemcell-resource/cmd/in/main.go

Lines 84 to 87 in 439cc17

	err = client.DownloadStemcell(stemcell, location, inRequest.Params.PreserveFilename)
	if err != nil {
	log.Fatalln(err)
	}