2021-11-10T01:20:26.467 runner[8d92de86] syd [info] Starting instance
2021-11-10T01:20:26.490 runner[8d92de86] syd [info] Configuring virtual machine
2021-11-10T01:20:26.491 runner[8d92de86] syd [info] Pulling container image
2021-11-10T01:22:26.683 runner[8d92de86] syd [info] Pull failed, retrying (attempt #0)
2021-11-10T01:24:26.786 runner[8d92de86] syd [info] Pull failed, retrying (attempt #1)
2021-11-10T01:26:26.890 runner[8d92de86] syd [info] Pull failed, retrying (attempt #2)
2021-11-10T01:26:26.890 runner[8d92de86] syd [info] Pulling image failed
The first node in the deployment started successfully but then the rest experienced the issue, making it worse due to some of the previous nodes being shutdown and the rollback unable to restore nodes.
If pull issues are breaking deploys (these are somewhat intermittent), you can get going by disabling auto rollback:
[experimental]
auto_rollback = false
If a VM fails during deploys, it leaves the rest in place. You can then run fly status to see outdated VMs, and try stopping them one by one with fly vm stop <id> to get them updated to the newest version.