This message was deleted.
# k3s
a
This message was deleted.
👀 1
l
What are you trying to achieve?
s
I'd like to develop in k3s on localhost in a containerized environment (docker). For this i use
<http://test.local.gd|test.local.gd>
which always resolves to localhost. I fixed internal k3s communication with coredns (although https fails). Problem: my k3s cluster is damaged when I restart it
I wonder why it's so difficult to develop without a public domain
c
fwiw, the docker-compose file is essentially unmaintained. k3d is a much better alternative if you want to run multiple k3s nodes in docker.
s
oh no 😒, i found advice on the internet that k3d is not needed anymore because k3s runs on docker just fine. I only need a single k3s node
and i wonder if k3d can be configured as code, just like docker compose
and how to make internal communication with https work on localhost subdomains
c
if you want to run just a single node, then yeah
docker run rancher/k3s
should work fine
s
it doesn't work
Copy code
failed to evacuate root cgroup: mkdir /sys/fs/cgroup/init: read-only file system
c
are you running an ancient version of k3s, or perhaps forgetting to run it --privileged?
s
yes i just did
docker run rancher/k3s
--privileged
works, now i get the error
--token is required
ah i need to write ...
server --token
...
and my favorite error is
Unable to get btrfs mountpoint IDs
Can k3d with a config file be a reasonable alternative to docker compose for single node? Will k3d be maintained in some years? https://k3d.io/v5.4.1/usage/configfile/#all-options-example
c
k3d is not our project, so I can’t speak as to its long term supportability
I haven’t personally used k3s on btrfs, but you might have to poke at that a bit. play with the --snapshotter option perhaps?
s
c
you might try adding the same option to k3s?
s
of course i did and the errors went away, however when a cluster destroys itself upon restart something is wrong
c
well yeah, you’re running it in docker. Did you want to create a persistent volume for the data? When you run something in docker, all the data that’s not on a volume goes away when you delete the container. that’s just how Docker works.
s
i know docker pretty well
why make a docker-compose file that does not preserve data? i don't get the logic
doesn't matter because it's unmaintained
• What fails: After the restart, you cannot connect to the cluster anymore and
kubectl
will give you a lot of errors • What causes this issue: it’s a known issue with dqlite in `k3s` which doesn’t allow the initializing server node to go down • What’s the solution: Hopefully, this will be solved by the planned replacement of dqlite with embedded etcd in k3s
c
we don’t use dqlite any more
haven’t for years