Add retry on server errror 502

[koppor]

In case of large projects, maven central response with 502.

See JabRef/jabref-koppor#686 for a non-MWE.

Cannot determine dependency age for "com.vladsch.flexmark:flexmark-util-html:0.64.8" and repository "MavenRepo" (reason: IOException: Response for URL https://search.maven.org/solrsearch/select?q=g%3A%22com.vladsch.flexmark%22%20AND%20a%3A%22flexmark-util-html%22 returned 502).

I think, a "simply" retry after 3 seconds should be done in the case if 502 is returned.

[f4lco]

Hello @koppor, thank you for your feedback. This project is accepting contributions, and your suggestion seems worthwhile. Would you feel confident implementing a retry for a status code of 502?

[koppor]

@f4lco Yeah, I feel confident. I put it on my TODO list after the (unrelated) issue openrewrite/rewrite#4054 :)

[Breefield]

+1 to this, I'm about to introduce this.
Would folks prefer an exponential backoff retry with total timeout, or just fail the entire report if 1x 502 surfaces?

I notice the 502s are not consistent and sometimes Maven is more agro than other times.

[f4lco]

@Breefield not sure if I get the question correct, let me put my 2ct that way: I think the point is to retry the same request for the same dependency at all.
I'd assume a fixed delay ("after x seconds") is fine, because we want to mitigate the effects of an unreliable network or server.
If we wanted to get around a rate limiter, we needed to find another way to throttle the requests or cache more aggressively.

Regarding the failure mode: I think the plugin already has the failOnError switch, which we can put to use:

• on failOnError = false, we activate "best effort mode", log the terminally failed request, and move on
• on failOnError = true, the entire report fails when all requests for a particular dependency have failed

@koppor what do you think?

[Breefield]

Oh yes, failOnError already exists.

This would just a be a retry policy with a retry limit.

[Breefield]

#37