Andrey Smirnov

Whether it's enough or not depends on your workload, and also these are VMs, so you can easily overcommit on memory.

This is your problem - it's either communication between etcd nodes, or disk I/O is too slow for etcd. It has nothing to do with Talos itself.

Troubleshooting depends on your environment/platform and debugging tools available with your environment.

> Did this get fixed? kube-router was updated to be less agressive about rpfilter, so it "just works" now. this PR hasn't been finished.

> If it works with kube-router now is this PR still needed? yes, otherwise we would have closed it enabling rp_filter by default would be nice as a feature

Applying same patch to the machine configuration repeatedly is not the supported gitops flow. See https://docs.siderolabs.com/talos/v1.11/configure-your-talos-cluster/system-configuration/reproducible-machine-configuration You can still achieve "replace" strategy by having two patches, one with `$patch: delete`...

> When applying such a patch (using the `$patch: delete` approach), we might temporarily break our production environment (depending on what changes are being applied) and then have to immediately...

Your patches are wrong, `$patch: delete` is at the wrong level, it should be attached to the element being removed, not to the parent element. Please see https://docs.siderolabs.com/talos/v1.11/configure-your-talos-cluster/system-configuration/reproducible-machine-configuration for a...

Please stop attaching screenshots where a codeblock would work. If you have questions on using patches, please open a discussion instead. your patch should look like: ```yaml machine: network: nameservers:...

I wonder if it's not related to swap at all, but rather any cold reboot doesn't work for you?