CLI needs to be resilient to network flakiness #763
Labels
area/cli
UX of using the CLI (args, output, logs)
impact/reliability
Something that feels unreliable or flaky
kind/enhancement
Improvements or new features
Milestone
We received a bug report from a user, who saw the following error message during an update to their stack:
error: waiting for update: calling API: performing HTTP request: Get https://api.pulumi.com/api/orgs/<org>/programs/<project>/stacks/<stack>/update/008694d1-c03d-4674-a29a-e04ca1352866?afterIndex=1513992748356531783: read tcp 192.168.1.22:49324->52.88.12.77:443: read: operation timed out
. Notably the192.168.1.22
address is the user's PC and52.88.12.77
is theapi.pulumi.com
endpoint.It would appear there was a flaky network problem that interrupted the querying of an update's status. We should add appropriate retry logic for this case explicitly.
Note that the ability to reattach the CLI to the current deployment would have come in handy here. Absent this, it appeared something truly catastrophic had occurred, and there wasn't much recourse.
The text was updated successfully, but these errors were encountered: