Ephemeral storage pdcsi node issue

[GianlucaStiga]

Hi, I have a lot of warning about pdcsi-node CPU consuming, but I don't know why and I can't assign them more resources, because they are automatically created and managed by GKE.

Can someone help me to understand this problem (if it is a problem..)?

thank you 

[Willbin]

Hello @GianlucaStiga,

Welcome to Google Cloud Community!

Seems like a process or a custom application that is running on your environment.

You have to determine which application, service or process is responsible for the 'pdcsi-node' CPU consumption. You can check the logs for the 'pdcsi-node' process to see if there are any error messages or warnings that might utilize your CPU resources.

You can post the error messages here if you find any, we are glad to help.

Thanks

[GianlucaStiga]

Hi Willbin,

do you have any suggestion about this pdci-node issue?

I have the same behaviour in another cluster, both have 1.25.10-gke.1200.

[GianlucaStiga]

Hi,

I have no recent logs on pdcsi-node, the last are the initialisation logs from 10 days ago, due to node restart. In the initialisation logs I found this warning, but seems not relevant: 

 This is my pdcsi-node pods status:

All my application pods status are ok (mem and cpu used are under the requested line).

I would like to know if I had to try to modify the automatically created pdcsi-node to add more cpu or if it is not a problem.

thanks

[jasperykj]

Hi all, 

Recently we also faced the same issue. PDCSI CPU usage spiked to 1000% causing all our servers to go down. 

[Istiaq]

hi, have you been able to resolve this?

[GianlucaStiga]

No. My processes keep going on, but the issue persists 

[CJ42]

I too have just noticed this issue on our cluster. Anyone find a root cause?

[wingrtjvcr · Accepted solution]

I resolved this issue after upgrading GKE to version 1.27.

[GianlucaStiga]

Thanks, I will try in my test environment. In production I can't, because I'm subscribed to gke "Stable channel", that is 1.26 at the moment