Does Forge still drop product events larger than 200kB?

Forge Forge Platform and Tools
, ,
1

The Forge product event docs mention at the very bottom of the page:

Product events larger than 200 kB are not delivered. This limit may change without notice.

Can anyone from Atlassian confirm that this limitation is still present?

I was looking at migrating our Connect webhooks to Forge, but if this is still the case, we are blocked. I double-checked with webhook data on our production Connect app and can see that we regularly get issue updated webhooks that are larger than 200kB.

1 Like
2

Hey Tobias, I have flagged this with the team owning this capability. They will respond to this soon.

Cheers
Ben

3 Likes
3

Hey Tobias,

Unfortunately yes, the limit still stands. Will update here as soon as we solve for this.

Thanks,
Rushikesh

3 Likes
4

Thanks for the update @RushikeshWalke. Based on your response, it sounds like this issue can be resolved and is not a fundamental system limitation. Would the fix still cap the payload size?

Also, do you have any indication of the timeline for a fix? Is there an issue we could watch?

Thanks!

1 Like
5

Hey Tobias,

There is a fundamental limit because of some tools we use, while I can not give you a timeline, I will look for an issue if created.

6

For Jira, there is a “Get failed webhooks” API.

  • Does it work with Forge or just for Connect webhooks?
  • Could webhooks not being delivered because of Forge queue size restrictions be added to this API as a short-term workaround? Or is this already the case?
2 Likes
7

I have created FRGE-1895 as I was unable to find an existing issue.

1 Like
8

Hi all,

To update you on this, I have been working with the owning team, they are looking at how to increase the current 200KB limit up to the new SQS limit of 1MiB (new limit will be slightly lower to make room for Atlassian overheads). Timelines on this are to be confirmed.

Based on the above shipping, the team will re-assess if any Product events are being dropped (we expect 0 to super low percentage from this point) and make further plans to close out any gaps found with the producers (Products) to guarantee message delivery.

Cheers
Ben

6 Likes
9

Hi Benny,

Thanks for sharing this info.
As I have been asked 100 times by Atlassian’s recently “Why did you not told us”, here is my take:
Any other result than 0 dropped events (not a small percentage) is a design failure and falls short for an extensibility framework of a multi-billion $ tech company. My hope would be that there are higher ambitions and you aim for the best-in-class event delivery framework and not something that just works (or event temporarily works until a product decides to send bigger events)

5 Likes
10

Hey @Benny @RushikeshWalke, could you please share an update on this issue? This is still one of the major blockers for our Connect to Forge migration.