K3s-server process

Hi Guys,

I just upgraded to the v2.4.5 everything looking good, but i have some minor thing that I’m curious of -

Why there is k3s-server process running with the Rancher on the same host? It uses twice amount of RAM and CPU is never lower than 6%


is s3-sever a typo? Do you mean k3s server?

This is my process list for a 2.4.5 rancher (launched via docker run … rancher/rancher, not via installing into a k8s cluster via helm)

root         1     0  0 16:38 ?        00:00:00 tini -- rancher --http-listen-port=80 --https-listen-port=443 --audit-log-path=/var/log/auditlog/rancher-api-audit.log --audit-level=0 --audit-log-maxage=10 --audit-log-maxbackup=10 --audit-log-maxsize=100
root         6     1  5 16:38 ?        00:00:03 rancher --http-listen-port=80 --https-listen-port=443 --audit-log-path=/var/log/auditlog/rancher-api-audit.log --audit-level=0 --audit-log-maxage=10 --audit-log-maxbackup=10 --audit-log-maxsize=100
root        18     6 13 16:38 ?        00:00:09 etcd --data-dir=management-state/etcd
root        34     6 99 16:38 ?        00:02:06 k3s server --no-deploy=traefik --no-deploy=coredns --no-deploy=servicelb --no-deploy=metrics-server --no-deploy=local-storage --disable-agent --datastore-endpoint=http://localhost:2379
root        66     0  5 16:39 pts/0    00:00:00 bash
root        83     6 65 16:39 ?        00:00:01 channelserver --config-key=k3s --url https://releases.rancher.com/kontainer-driver-metadata/release-v2.4/data.json --url=/var/lib/rancher-data/driver-metadata/data.json --refresh-interval 24h0m0s --listen-address= --channel-server-version v2.4.5 v2.4.5
root       100    66  0 16:39 pts/0    00:00:00 ps -ef

Yes, s3 was a typo - fixed.

My Rancher is launched the same way (docker run) and the cluster is the RKE one (got imported).
Here is the screenshot:
Screen Shot 2020-06-22 at 10.17.42 AM

Seems like you have the k3s-service process running as well, what is it for and why it is so busy?

Rancher is kubernetes application. It needs to run on top of kubernetes cluster. If you do our “HA” install using helm, it’ll use the cluster that you are deploying it into. If you launch rancher via docker run, the container spins up a k3s server process to act as the kubernetes cluster.

That’s what that process is and it can be busy because it is doing all the kubernetes things, like reading from and writing to the DB for any api calls that rancher needs to make (which can be a lot if you have a lot of clusters/nodes/other resources being managed by Rancher).

Makes sense, Thanks for clarification.