Issue: Instance upgrade intermittently fails with `UPGRADE_STREAM_ERROR` but succeeds on retry

[sunglow666]

Summary

Upgrading an agent instance from 0.27.0 to 0.28.0 failed on the first attempt with a generic stream error, but the same upgrade succeeded after retrying.

---

Affected instance

• Instance name: warm-cow-fasuh
• Upgrade: nearaidev/ironclaw-dind:0.27.0 -> nearaidev/ironclaw-dind:0.28.0
• Instance ID: f06dab3f-cf7a-492b-952a-34bc4717ed2e
• Environment: nearai-prod

---

API response

Endpoint:

POST https://private.near.ai/v1/agents/instances/f06dab3f-cf7a-492b-952a-34bc4717ed2e/upgrade

Event stream response:

event: image_changing
data: {"event":"image_changing","name":"warm-cow-fasuh","oldImage":"docker.io/nearaidev/ironclaw-dind:0.27.0","newImage":"docker.io/nearaidev/ironclaw-dind:0.28.0","env":"nearai-prod"}

: keepalive

: keepalive

data: {"code":"UPGRADE_STREAM_ERROR","error":"Upgrade failed"}

---

Observed behavior

The first upgrade attempt failed with:
UPGRADE_STREAM_ERROR: Upgrade failed

No more specific error message appeared to be exposed in the UI or event stream.
After retrying the same upgrade, it completed successfully.

---

Additional reports

This does not appear to be isolated to a single instance:

• Another one of my agents, glad-pig-nodoj, showed the same behavior.
• Another user also reported a similar failure:
  Upgrade stream error: instance_id=f9696fa3-542a-4b73-88b7-a85b10d0f65b, error=error decoding response body