I noticed some odd behaviour when a node is starting afresh and lots of containers are being scheduled: it will often error out and abort with odd iptables error messages. It turns out this is due to not sharing the node's global xtables lock.
Unfortunately it's not as simple as just passing through the lockfile: by default iptables doesn't wait for the lock and instead just returns an error when the lock is held. So this PR also updates iptables.go to correctly wait for the lock if it is held.
I noticed some odd behaviour when a node is starting afresh and lots of containers are being scheduled: it will often error out and abort with odd iptables error messages. It turns out this is due to not sharing the node's global xtables lock.
Unfortunately it's not as simple as just passing through the lockfile: by default iptables doesn't wait for the lock and instead just returns an error when the lock is held. So this PR also updates
iptables.go
to correctly wait for the lock if it is held.