Wait until nodes connected before infra upgrade #122
Conversation
Previously, it was observed several times that nodes might be in a state of disconnection from the orchestrator when peadm tries to perform a `puppet infra upgrade` operation. This commit implements a wait function to ensure that all nodes are orchestrator-connected before beginning a puppet infra upgrade operation.
| http = http_object | ||
| request = request_object(nodes: nodes, token_file: token_file) | ||
| inventory = {} | ||
| Timeout::timeout(timeout) do |
There was a problem hiding this comment.
In this article: https://medium.com/@adamhooper/in-ruby-dont-use-timeout-77d9d4e5a001 the dangers of Timeout::timeout are outlined. Is the worry that this code might raise for unexpected conditions justified?
There was a problem hiding this comment.
In this code, we only care about the assignment operation to the variable inventory. The objective isn't to hard-lock execution time to a particular interval, it's just "stop trying after n seconds if you haven't succeeded". As such, I believe the behavior of Timeout here will not cause anything unexpected to happen. The worst-case scenario is that a hail-mary last-second possible success might not be counted if the code is interrupted after making an inventory request but before checking the results. That's okay. If it took that long to get a successful result, well, we exceeded the timeout. 🤷
Previously, it was observed several times that nodes might be in a state of disconnection from the orchestrator when peadm tries to perform a
puppet infra upgradeoperation. This commit implements a wait function to ensure that all nodes are orchestrator-connected before beginning a puppet infra upgrade operation.