Sudden spike in Forge Functions Usage and charges

Forge
1

I am seeing a sudden spike today in the Forge Functions Usage and charges for two of our apps even though nothing changed and no new deployments for a few days.

Has something changed in the way Forge tracks usage and pricing for functions?

2

Hi @GirishReddy , there has been some changes to the usage calculations for async functions and scheduled triggers. Please see: https://developer.atlassian.com/platform/forge/changelog/#CHANGE-2988

3

Thank you @BoZhang Can you provide more details on what changed?

We do use async functions and scheduled triggers but the number of executions should not be as high as being reported in our console metrics dashboard.

How can we get a bit more granular information so we can understand what’s going on?

4

Hi @GirishReddy, there has been an update around this, please see: https://developer.atlassian.com/platform/forge/changelog/#CHANGE-2992

2 Likes
5

Thanks @BoZhang - appreciate the update and the action to revert the fix.

Will there be any additional docs on how async functions and scheduled triggers are computed? From what I can tell our number of executions should not be as high as being reported in our console metrics dashboard. How can we get a bit more granular information so we can understand what’s going on?

6

Hi @GirishReddy, the team will be re-evaluating how we proceed with async events & scheduled triggers. Please follow the Forge changelog, updates on this topic will be provided there when available.

7

Hello @BoZhang

We also observed an increase in the forge compute metric during the same period, without any change on our side. Since the rollback, metrics are back to normal.

We would appreciate a clearer explanation of how compute will be calculated once the fix is restored.

Most importantly, we need a way to forecast usage under the restored calculation method. today we cannot see the “fixed” metric, so we cannot validate the optimizations we implement or confirm that our changes effectively reduce consumption before the fix returns.

A solution that would work for us would include:

  • A preview mode or sandbox metric showing the “new” compute calculation alongside the current one

  • A documented formula with a reference example (sample inputs and expected outputs)

Thanks

8

Hi @christophe.prome, as part of the observability, we will show the async specific compute usage in the developer console for few months prior starting to charge.

1 Like
9

Is it possible to add conditions to scheduled triggers before this change? We did see a 10x increase in Functions usage due to this change which is likely caused by our 5 min scheduler which only exists because there is no reliable way to implement install / upgrade task

1 Like