Home / server / Questions

Questions[portworx](server)

Martin Hope
mr.zog
Asked: 2022-01-16 17:07:28 +0800 CST
  • 0

I have seven kubernetes nodes, two of which are physical (bare metal) servers. One of these physical machines has a SSD I'd like to use as a portworx cache disk. Is this a supported configuration or will kubernetes endlessly look for that one disk on every host? Here's a snippet from the "spec" file I am using.

  containers:
    - name: portworx
      image: portworx/oci-monitor:2.8.1.2
      imagePullPolicy: Always
      args:
        ["-k", "etcd:https://kube-kvdb.nppl.yo:2379", "-c", "px-nppl-00dZ16f4-1e2c-4afd-97fa-gf5aef26f24e", "-s", "/dev/vdb", "-cache", "/dev/sdb", "-secret_type", "k8s",
         "-userpwd", "$PX_KVDB_USERNAME:$PX_KVDB_PASSWORD",
         "-x", "kubernetes"]

And here is the state of my cluster.

msh@kube-kvdb:~$ kubectl get pods -n kube-system -o wide |grep port
portworx-5n9fx                         1/2     Running            21 (8m39s ago)   5h36m   192.168.10.4    landrew     <none>           <none>
portworx-api-66z6c                     0/1     Running            3 (158m ago)     5h36m   192.168.10.4    landrew     <none>           <none>
portworx-api-cx5k2                     0/1     Running            1 (126m ago)     5h36m   192.168.10.35   kube1       <none>           <none>
portworx-api-h6cj6                     0/1     Running            42 (135m ago)    5h36m   192.168.10.36   kube2       <none>           <none>
portworx-api-ngj9j                     0/1     Running            0                5h36m   192.168.10.34   kube0       <none>           <none>
portworx-api-zcdd8                     0/1     Running            0                5h36m   192.168.10.37   kube3       <none>           <none>
portworx-cnpt2                         1/2     CrashLoopBackOff   74 (93s ago)     5h36m   192.168.10.35   kube1       <none>           <none>
portworx-kj7vs                         1/2     CrashLoopBackOff   114 (42s ago)    5h36m   192.168.10.36   kube2       <none>           <none>
portworx-lf9l6                         1/2     CrashLoopBackOff   69 (3m28s ago)   5h36m   192.168.10.34   kube0       <none>           <none>
portworx-t6s4f                         1/2     CrashLoopBackOff   69 (3m54s ago)   5h36m   192.168.10.37   kube3       <none>           <none>
kubernetes portworx
Martin Hope
mr.zog
Asked: 2022-01-10 11:02:13 +0800 CST
  • 1

I have a tiny (three worker node, one controlplane) kubernetes cluster running and I want to configure portworx for storage. All my hosts are kvm virtual machines running Ubuntu 20.4.3 LTS with 5.4.0-73-generic kernel. I used https://central.portworx.com/specGen/wizard to generate a portworx 2.8.1.2-c9b840b (OCI) spec file and applied it to my cluster. I chose to not configure a dedicated kvdb drive on these hosts although each host does have a sdb intended as the portworx drive. (These sdb drives are not mounted.)

When I run pxctl status I see:

NODE NodeStartFailure 3a5bce08-b8ed-4fca-93a2-435a8d6e512b ALARM 1 Jan 9 18:31:45 UTC 2022 Jan 9 18:31:45 UTC 2022 > Could not find any available storage disks on this node. Node cannot act as a kvdb node as no storage device found. Waiting for other nodes to start kvdb. > If this is a storage node please add disks and restart Portworx.

NODE InternalKvdbSetupFailed 907d5a2c-297d-4587-87f8-31d539cff440 ALARM 1 Jan 9 18:31:10 UTC 2022 Jan 9 18:31:10 UTC 2022 failed to setup internal kvdb: failed to provision internal kvdb: timeout waiting for other kvdb nodes to start: Could not find any available storage disks on this node. Node cannot act as a kvdb node as no storage device found. Waiting for other nodes to start kvdb. If this is a storage node please add disks and restart Portworx.

I would rather use etcd instead of kvdb. Does anyone care to comment? Also, is the portworx operator intended for kubernetes?

oops. I just noticed this on the portworx website:

PX-Essentials support only built-in KVDB.
External ETCD is accessible in PX-Enterprise

Thanks.

kubernetes portworx