I’ve just performed a bare metal install of R2.5.3 on top of RO1.5.6 all completely default with no workloads installed. I can hear the hard drive constantly poppling now that R2.5.3 is installed, can anyone tell me what it is likely to be due too and what I can do to fix it ?
I am a Linux/ Docker/ Rancher newbie so any tips on what I can do to diagnose this woukd be much appreciated.
Anyone got any ideas on this ?
Kubernetes has lots of things with status or heartbeats reporting more or less constantly. Each change results in a write to the database. Each write to the database has to be synced to disk to provide the guarantees it does. Even an empty cluster is still doing most of the same overhead.
Running etcd on a spinning disk is also not particularly recommended because they are so much slower and higher latency versus a SSD.
Well, I was just passing by and thought that I’d post the answer:
It was docker that was logging and there seems to be no way to prevent an existing container from doing that if it was initially set up to do so, therefore I eventually re-created my rancher installation as follows :
docker run -d --privileged --log-driver none --restart=unless-stopped -p 8080:80 -p 8443:443 rancher/rancher:latest
The –log-driver=none is required to prevent docker logging from running.
I think the advice given above to run on an ssd when the logging is constantly writing to it may be poor for some ssd devices as they may suffer from write hardening under those circumstances.
Hope this helps somebody,