Hi everyone,
I created this github issue but I have the feeling the repository is quite inactive so I thought to try my chance here.
I have some performance issues when using cvmfs from a pod using the cvmfs-csi. In the link you can see a table where I compare some commands. You can see a significant drop in performance when using cvmfs from the container than from the VM.
If I compare this to my setup without the cvmfs mount, this is also a significant performance drop.
Anyone has clue why this sudden drop in performance. Is there something essential that is missing in this setup?
Mohammad
Hi Mohammad,
welcome to the forum! That’s a very interesting question: some overhead is to be expected, but it should not be significant. We’ll try to reproduce this and investigate.
Great! Let me know if you need some more details. Most of it you will find in the github issue.
I have been monitoring the resources for the pods deployed by the daemonset, apparrently it was CPU throttling with these assigned resources. Increasing the resources (upto a core for the CPU limit) on the relevant containers (automount and automount reconciler) did resolve my slow responsiveness from the CVMFS mount.
What I learned:
Depending on your workload, you will need to set the resources for the containers in the cvmfs-csi plugin accordingly.