Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
retry various network create errors (#161)
"hardcode" a capped backoff retry for network create requests. while the daemon does its best to find a non-conflicting cidr range to create a network on, when multiple concurrent requests are trying to create a network (such as when multiple harnesses are being created), theres a relatively common case where conflicts happen. a "dumb" retry like this seems to be common place in other solutions like [`kind`](https://github.com/kubernetes-sigs/kind/blob/0f1c56884daf22bd4d1559f9b955c26afaaeefb5/pkg/cluster/internal/providers/docker/network.go#L84), so this approach just adds a non-configurable retry loop with a reasonable backoff and cap. we could get more fancy and predetermine a free cidr space during a locked op, but that feels unnecessary for now.
- Loading branch information