Cloud agent is stuck

Support Bug Reports
,
1

Where does the bug appear (feature/product)?

Background Agent (GitHub, Slack, Web, Linear)

Describe the Bug

Today i was delegating some issues to cursor from linear and both cloud agents are stuck. Restarting didn’t help

Steps to Reproduce

  • ask cursor agent to implement issue from linear

Expected Behavior

  • same as it was before - PR created ready for review

Screenshots / Screen Recordings

1772775829831.jpg
1772775829831.jpg776×335 39 KB

Operating System

MacOS

Version Information

Cloud agent

For AI issues: which model did you use?

Opus 4.6

For AI issues: add Request ID with privacy disabled

Agent id: bc-a73651f9-4a8e-48c6-b7a6-ecd7558130b0

Does this stop you from using Cursor

Yes - Cursor is unusable

3

Hey, thanks for the report. The agent uploaded the snapshot to 100% (2.31 GB) but got stuck on “Setting up repository”.

While we’re looking into it, a couple questions:

  • Does this happen with every new agent you start from Linear, or only with these two?
  • If you start a cloud agent directly from the UI (not from Linear) on the same repo, does it get stuck too?

Let me know how it goes.

4

Well, I am just started working, so these two got stuck. I think next one will stuck as well.

Tried to start from cloud agents web page (copy pasted prompt from stucked one) - seems executing.

5

@deanrie

tried to start second one from the web ui - and seems this got stuck now.. first one completed the task.

image
image802×158 11.5 KB

id of stuck one: bc-526e7a20-b304-4de4-a6c4-795c128b9010

6

stopped it, start another one - and it seems executing

not clear what causing this, either running in parallel, or smth else

anyway, super annoying and not reliable, since i have to sit and monitor whether it is working or not

7

I see, it looks like the issue happens when the second agent tries to use a snapshot that the first agent is still building, or just finished building. That would explain why one runs and the other gets stuck on “Waiting for snapshot”.

For now, as a workaround, please run the agents one after another, not in parallel. Wait until the first one gets past the “Setting up repository” stage, then start the second.

A couple more questions:

  • Are you using private workers for cloud agents, or the default infrastructure?
  • Is the repo size always the same and always the same repo, or different repos?

Let me know if the issue still happens even with sequential runs.

8
  • we are using default infra
  • in this case it was the same repo (and usually it is same repo for us)
9

Thanks for the info, the picture is clear now. The problem is that parallel agents on the same repo are conflicting over the snapshot. The second one tries to use a snapshot that the first is still building, or has just finished building.

For now, running things sequentially is the only reliable workaround. I’ll let you know if there’s any news on a fix.

12

This topic was automatically closed 22 days after the last reply. New replies are no longer allowed.