What are the http codes to automatically retry the request?

Question

I am using the mix Spring-Cloud + feign + spring-retry to help retry requests on the client side (all are Kotlin-based back-ends)

My spring-boot conf is like this:

myApp:
  ribbon:
    OkToRetryOnAllOperations: true
    retryableStatusCodes: 404, 503

(note: OkToRetryOnAllOperations=true is only present to retry also POST/PUT requests)

Retrying 404 & 503 HTTP codes sounds good, but I cannot figure out if there is a "classic" or "default" list of error codes to retry. Does this kind of good practice exist?

We suppose all the requests are idempotent on the server side (if not, retrying could cause problems).

Those codes could be relevant to retry the requests too: 408, 500, 502, 504 ? — Jeremy L, Aug 09 '18 at 14:55
Does this answer your question? [Which HTTP errors should never trigger an automatic retry?](https://stackoverflow.com/questions/47680711/which-http-errors-should-never-trigger-an-automatic-retry) — Michael Freidgeim, Mar 03 '23 at 12:27

Alexey Soshin · Accepted Answer · 2023-03-20T21:34:29.663

19

As a very rough rule of thumb:
4XX - client did something bad
5XX - server did something bad

But it very much depends of the actual API.
Should you retry 500? Maybe, because the server had an unexpected hiccup while connecting to DB. Or, maybe you're sending it something it's not expecting, and instead of returning you 4XX it crashes.

There is usually not much reason to retry 404, unless you expect that this resource will appear.

The only HTTP codes which is valid for retry are 408, 502, 503 and 504

AWS clients also retry 429.

edited Mar 20 '23 at 21:34

answered Aug 09 '18 at 15:05

Alexey Soshin

16,718
2
31
40

Thank for your response. I thought that 404 was relevant in the case of network problems on the client side, don't you think? – Jeremy L Aug 09 '18 at 15:42
3

No, because those are the errors returned by the server. The fact that you got 404 means that you managed to reach the server. – Alexey Soshin Aug 09 '18 at 15:44
1

you're right, if the resource does not exist, there is no chance that a retry makes it appear... thank! – Jeremy L Aug 10 '18 at 08:53
6

429s are also retryable. – sbenderli Jun 15 '21 at 20:22
1

Only retry 429s if you know what to wait for. You're creating undue load and your retries may deplete rate limiting quota continuously. – Robert Hensing Aug 23 '21 at 10:36

score 10 · Answer 2 · answered Nov 30 '22 at 11:55

10

In general, I recommend these:

408 Request Timeout
425 Too Early
429 Too Many Requests
500 Internal Server Error
502 Bad Gateway
503 Service Unavailable
504 Gateway Timeout

But it really depends on the API, so you can look up its documentation if it has it. For example, Whatsapp uses 500 for errors that can be retried or not.

answered Nov 30 '22 at 11:55

Lucas Vazquez

1,456
16
20

But should 418 be retried? – Morgan Touverey Quilling Apr 27 '23 at 15:40
3

not, but you should have your cup ready – Lucas Vazquez Apr 29 '23 at 19:51
Yay for the answer. Boo to whatsapp for making everything 500. – granadaCoder Jul 07 '23 at 15:08

What are the http codes to automatically retry the request?

2 Answers2

Linked