view: hack attempting to mitigate split-brain GRPC failures

[hdevalence]

This isn't a permanent fix, and it's not the right way to solve the robustness issue (more details TK) but it may help with the immediate problem with summonerd, and seems unlikely to add new failure modes.

[cronokirby]

What's with the protobuf lint failures?

[hdevalence]

I think it's comparing the state of this branch against main, even though the PR is being made against a different branch. Probably the solution is to change the CI workflow to filter to only apply to PRs against main.

[cronokirby]

Also, are we sure that one second is enough?

[hdevalence]

Also, are we sure that one second is enough?

We are not! But it seems likely, and fits in the middle of the confirmation delay we already put in pcli. If it isn't, we'll see it really clearly in the logs, because there will be a warning before an error. So I think it's probably a reasonable starting point.

[cronokirby]

Also, are we sure that one second is enough?

We are not! But it seems likely, and fits in the middle of the confirmation delay we already put in pcli. If it isn't, we'll see it really clearly in the logs, because there will be a warning before an error. So I think it's probably a reasonable starting point.

Yeah, makes sense, if we still run into this we could do a few retries with exponential backoff, but I'm in favor of merging this as an obviously correct patch for now