Geek Answers Handbook

Can ReadWriteOnce PVC be rejected by AWS EBS / GCP PersistentDisk when rescheduling subdomains?

Suppose I currently have 2 PVCseconds ReadWriteOnce, approved Pod #1and Pod #2accordingly. Both of them work on Node #1.

Next, Pod #2updated with a new Docker image. However, at the same time it Pod #3starts and stands out Node #1. Since it is Node #1now full, Pod #2only Kubernetes can be allocated on Node #2.

Since AWS EBS and Google PersistentDisk could only be installed on one Node, could you not Pod #2connect to the previously announced PVC?

If so, how to avoid this problem?

+6
source share
2 answers

Yes, this is the drawback of pv / pvc with current AWS and GCE storage shipments.

To avoid this, you will have to use a different storage infrastructure that does not have this limitation. CEPH, Gluster, scaleIO (and others) are possible. These solutions abstract storage from disks and provide a storage tier that is no longer dependent on the node.

+2
source

This should not be a problem. When Pod #2scheduled for Node #2, Kubernetes should automatically separate the volume from Node #1and attach it to Node #2for Pod #2to use there.

+1
source

More articles:


All Articles