Is it possible that Jira automations fail randomly?

[Jonathan Berneman]

Our teams have several automation rules and lately we have been having some of them failing. The failure is not on how the rule is written because it actually says in the log that the edit has been done, yet the issue doesn´t show the change. 

For example: When "X" creates an issue, a label "X project" is added.
The rule is executed with no problem and the log seems perfect (including the statement that the editing happened) but then the issue doesn´t show. 

And in this case its happening that there are 3 labels to be included and maybe it says all three have been included but then only one shows up in the issue. 

Last but not least, it happens randomly. Sometimes it edits two of the three labels, sometimes all three, sometimes none.

We have already a lot of experience with automations and we have already checked if the rule was written incorrectly, if  there were any permissions issues but the rule is fine and within the same issue it can happen (the failure) randomly. 

[John Funk]

Hi Jonathan - Welcome to the Atlassian Community!

I don't think that automation rules fail randomly. There is always a reason - and that should show in your audit log for the rule. If it doesn't run at all and you think it should, then make sure the box is check on the rule Details page for Allow trigger rule. 

[John Funk]

Hey @Jonathan Berneman 

Did you ever get this figured out? 

[Jonathan Berneman]

Hi @John Funk not really.
There were actually some failures on the automation part which weren´t a problem on how it was set up. 
The issue is the chronology on which it executes certain automations. 
So for example if the automation will edit the labels field by first eliminating the current labels (in this case there were two labels), and then add two different labels, then the execution of the automation actually would have to do steps like eliminating one, eliminating the second one, then adding one, then adding the other one. 
And in those smaller execution steps, it would sometimes fail due to arbitrary order in which it did that and other rules executing like some rules that would trigger if that field was empty (so it would end up stopping in the middle of the process and we ended up with one of the two labels or sometimes both labels added but not the previous two eliminated). And I say its random because it wasn´t always the same arbitrary order and there was something we could do inside that detail.  
We ended up changing some automations to avoid this but it did implicate a lot of work and not being able to use it to a full extent.
We followed it with a person from Atlassian helping us but realising that there were some thing that needed to be improved in the automations way of working

[John Funk]

Thanks for the further explanation. Maybe implementing some re-fetches between steps would help as well. 

[Fabian Lim]

Hi @Jonathan Berneman

Welcome to the community!

Maybe you have done this, but check the ticket's history as someone may have removed the labels by mistake.

The rule is very simple. I would also consider contacting atlassian support directly to see if they find any issues in your environment.

Regards