Process to run VDBench Raw Volume benchmarks on Portworx Volumes
This assumes you Already have a working Portworx by Pure Storage cluster in place. If you do not, visit https://portworx.com for more information.
By default Kubernetes CSI provisioned volumes are consumed with a file-system in place. However, the timeless benchmark utility, VDBENCH, frequently comes up in conversations with those evaluating data management solutions for Containers.
Here you will find a set of YAML files to provision 3 Portworx Volumes, a manual step to mount them to a host, and a YAML file to bring them up in a Kubernetes managed pod running VDBench. I have also included sample tests in a ConfigMap. This can be edited to change the testing parameters.
First you will want to download and run vdbench-raw-pvc.yaml. This will create 3 PVCs to use for the benchmarking. You can verifiy their creation by issuing:
kubectl get pvc
You should see your three volumes listed.
Next, SSH into the host you want to run the benchmarks from and issue the following commands:
pxctl volume list
Note the unattached volumes in the list. There should be 3 and these are the volumes you created previously. In the results of that command, you will want to copy the Name or ID of each volume in the next commands. Here is a sample output:
ID NAME SIZE HA SHARED ENCRYPTED PROXY-VOLUME IO_PRIORITYSTATUS SNAP-ENABLED
1054773657008748565 pvc-1a8130f4-98cf-4ef1-9bbb-f293473ccec3 50 GiB 2 no no no HIGH up no
234970879122504832 pvc-3834f418-9a45-42da-94a2-f69f7afb9131 50 GiB 2 no no no HIGH up no
926217724212074340 pvc-3b60289d-f633-4e9e-97f3-c77d32d5508a 50 GiB 2 no no no HIGH up no
966505478598664814 pvc-4a04bb92-a8c8-43f9-ad75-2a7700df6321 10 GiB 1 no no no HIGH up - attached on 10.10.100.102 no
336691233451103348 pvc-683d136a-b9cf-4dc2-8c0e-dc6f7273cce9 50 GiB 1 no no no HIGH up - attached on 10.10.100.102 no
985137598051312616 pvc-6ab79f32-def9-44af-b35b-474a38b006a8 5 GiB 1 no no no HIGH up - attached on 10.10.100.102 no
107592820124132570 pvc-d5e878f6-ce7b-44e2-a1cb-6342859d75c3 10 GiB 1 no no no HIGH up - attached on 10.10.100.102 no
To mount the unattached volumes to the host you need to copy the ID of the three volumes and issue:
pxctl host attach 1054773657008748565 && pxctl host attach 234970879122504832 && pxctl host attach 926217724212074340
This will mount the 3 volumes as /dev/pxd/pxd1054773657008748565, etc. You can issue >ls /dev/pxd to verify the are mounted. With the name of the 3 volumes copied to a text editor or some other means, you then log out of the host and proceed.
Next download vdbench-block-local.yaml and edit the device paths in that file to match those on your host. Also replace any instances of nodename with the hostname of the node you mounted the volumes on.
Once you have finished your substitutions, you can issue kubectl create -f vdbench-block-local.yaml to start the pod. Once the pod is created, Exec into it. VDBENCH is located at /vdb/vdbench and the job configuration will be in /templates/vdbench.job. So, to run the benchmark you can run:
kubectl exec -it pod/vdbench-rawblock -- /bin/bash
This will place you in the root of the pod.
./vdb/vdbench -f /templates/vdbench.job
This will kick off the VDBench job and display the results.