How to retry logic for up to N minutes?

Hi,

I would like to write a policy to retry my logic until it succeeds or until a “retry timeout” happen (not a “logic timeout”). With Policy.Handle<Exception>().Retry(3) I can tell Polly to retry up to 3 times. What I would like is to Policy.Handle<Exception>().Retry(TimeSpan.FromMinutes(5)).

Is there a clean way of doing this using Polly API? Thanks!

[reisenberger]

Hi @darxis !

• TimeoutPolicy provides the ability to time out any execution.
• PolicyWrap provides the ability to compose multiple policies into a single policy for later use.

For your Use Case, wrap a timeout policy around any retry policy you like. For example, the following should retry with 10 second delays between tries, up to a 5-minute limit:

Policy timeoutAfterFiveMinutes = Policy.Timeout(TimeSpan.FromMinutes(5));
Policy retryEveryTenSeconds = Policy.WaitAndRetryForever(iteration => TimeSpan.FromSeconds(10));
Policy tryEvery10SecondsUpTo5Minutes = timeoutAfterFiveMinutes.Wrap(retryEveryTenSeconds);
// You could create the above all in one statement.  I used long-named interim variables just to make the components clearer.

// Usage
tryEvery10SecondsUpTo5Minutes.Execute(...);

(EDIT: For a concise syntax, you can of course create that compound policy with an extension method.)

Comments to this stackoverflow question include various working dotnetfiddle examples and discussion of further nuances. They are async examples, and for much shorter timescales (because dotnetfiddle limits overall execution time), but the principles are the same.

Let us know if this gives you what you need!

[TylerNielsen]

I had a follow up question to this thread. If I’m reading the documentation on Timeout Policy correctly, this may be slightly different than what some users may expect based on OP ask.

Based on the Timeout Policy documentation, it seems that Timeout Policy will cause a synchronous task to be put on a separate thread and abandoned if the timeout is hit. I suspect OP might have been asking for a Policy that would simply not try again if timeout was reached, but not necessarily “walk away” from the executing process if it’s still going. If OP wasn’t interested in this behavior, I am 😃.

Would there be anyway to let the execution finish but simply not start another retry if the timeout had expired?