John Howard
John Howard
``` net.ipv6.conf.all.disable_ipv6 = 1 net.ipv6.conf.default.disable_ipv6 = 1 net.ipv6.conf.eth0.disable_ipv6 = 1 net.ipv6.conf.lo.disable_ipv6 = 1 ```
+1, we have a similar use case of our proto wire format being hashed to compare equality where this would help. I do wonder if we should just modify "strict"...
`failed to start proxy: Bind([::1]:15053, Os { code: 99, kind: AddrNotAvailable, message: "Cannot assign requested address" })` Not sure if this would cause the issue entirely, but it might. This...
Yeah I am looking into it
Arguably its a bug in EKS to not provide IPv6 support when the node starts as well... though we should have retries
Hmm, in my own testing we do retry. But the logs shown are weird: ``` info inpod::workloadmanager handling new stream info inpod::statemanager pod received, starting proxy uid="4a7f439b-b4c0-4f11-b6af-7d4bcb4563b0" name="stackstorm-ha-st2actionrunner-5d9fc664c7-fr49p" namespace="worker" info...
Cannot assign address is fixed in 1.23.1 which is out in a few hours
Great to hear. Thanks!
The logging never logged retries. What was probably happening is you have 2 proxies here. So the client retries to the egress GW, so you see 3 logs. There are...
So probably it's not retrying at all.