Workflow Failure Recovery

[dennissetiawan]

Hello , i have cases where I limit the duration/attemp of a state execution and I want to do rollback on my data in case of workflow failure to fix the data. Like a defer / recovery function in golang of some sort? Can we do this in iwf / what is the recommendation to do this?

[longquanzheng]

A common pattern that we are using is:

• put the logic in waitUntil with WaitUntilApiFailurePolicy.PROCEED_ON_FAILURE . It can just return empty commandRequest after the logic.
• This failure policy (in state options) will let the state continue to execute api after waitUntil has maxed out attempt/duration, rather than failing the workflow
• If waitUntil succeed, then continue the workflow as normal
• If waitUntil fails, then do a cleanup/recovery as you need before failing the workflow .

To check if waitUntil has succeeded, there is a field in CommandResults.getWaitUntilApiSucceeded()

[longquanzheng]

See the docs https://github.com/indeedeng/iwf#waituntil-api-failure-policy

[dennissetiawan]

Okay, so we can execute the recovery code in Execute API after set the policy if the GetWaitUntilApiSuceeded return false right?
when does that returns nil (in go SDK) StateWaitUntilApiSucceeded *bool.

Also what if the code i want to recover from executes in the Execute() API function?
Is there a way or i should just move it to another state?

[longquanzheng]

It will not return nil unless your server is very old. This was for a compatibility for the very old sever

[longquanzheng]

Yes another way is to utilize the “attempts” or “firstAttemptStartTime” from the context, and make sure it don’t use up the attempts , and return a decision to another state to cleanup/recover

[longquanzheng]

For execute api, ideally we can do this
#311

But it’s not prioritized yet . We are working on adding more resources to this project hopefully this will come later