Hi All,

I am trying to install MAS on a GCP cluster (3 master 3 worker), at the very end of the installer when it is trying to create the installer pipeline, it is creating a PVC named config-pvc under the mas-{instanceid}-piplines and requesting 500MB worth RWX capable storage. Thing is the only option is GCP capable of RWX is filestore and min 1TB needs to be provisioned which seems unnecessary given these pipelines only need to be run once. Any workaround for this issue?

Abhinav

Hi Abhinav,

You're right, the choice of ReadWriteMany (RWX) forces you to use such an overkill as filestore. We faced the same challenge and did not want to go this route. What we did instead - switched to ReadWriteOnce (RWO) access mode during the MAS installation:

mas install ... --storage-accessmode 'ReadWriteOnce'

Standard persistent disks in GCP support RWO mode, the PVC will bind to an RWO persistent disk without filestore. While MAS installer pipelines default to RWX for config-pvc in the mas-{instanceid}-pipelines namespace, this runs once and typically doesn't require multi-node access.

Hi All,

I am trying to install MAS on a GCP cluster (3 master 3 worker), at the very end of the installer when it is trying to create the installer pipeline, it is creating a PVC named config-pvc under the mas-{instanceid}-piplines and requesting 500MB worth RWX capable storage. Thing is the only option is GCP capable of RWX is filestore and min 1TB needs to be provisioned which seems unnecessary given these pipelines only need to be run once. Any workaround for this issue?

Abhinav

Hi @Ivan Lagunov

Thanks for the reply, I was able to complete the install yesterday by forcing config-pvc in RWO mode. On a separate note since you have installed MAS on GCP, my cluster is backed by balanced-pd's, How is the performance overall, any performance related issues?

Also my DB resides outside the cluster and plan to use balanced-pd's for the same. Not sure if they are good enough for a Maximo DB (these are supposedly backed by SSD's)

Again apologies for asking too many questions

Abhinav

Hi Abhinav,

You're right, the choice of ReadWriteMany (RWX) forces you to use such an overkill as filestore. We faced the same challenge and did not want to go this route. What we did instead - switched to ReadWriteOnce (RWO) access mode during the MAS installation:

mas install ... --storage-accessmode 'ReadWriteOnce'

Standard persistent disks in GCP support RWO mode, the PVC will bind to an RWO persistent disk without filestore. While MAS installer pipelines default to RWX for config-pvc in the mas-{instanceid}-pipelines namespace, this runs once and typically doesn't require multi-node access.

Hi All,

I am trying to install MAS on a GCP cluster (3 master 3 worker), at the very end of the installer when it is trying to create the installer pipeline, it is creating a PVC named config-pvc under the mas-{instanceid}-piplines and requesting 500MB worth RWX capable storage. Thing is the only option is GCP capable of RWX is filestore and min 1TB needs to be provisioned which seems unnecessary given these pipelines only need to be run once. Any workaround for this issue?

Abhinav

Hi Abhinav,

The short answer is that balanced persistent disks generally should be a solid choice. The long answer is that it depends on the sizing of the cluster and the expected load, so I advise you to monitor the load and change if it becomes the bottleneck. Most often performance issues are caused by the database or application bugs.

As for us, we actually tried even standard persistent disks as they were allocated for standard-csi storage class in our case in MAS 8.11. I'd not advise using standard persistent disks as they are backed by HDD but balanced persistent disks is a good option to start with.

Hi @Ivan Lagunov

Thanks for the reply, I was able to complete the install yesterday by forcing config-pvc in RWO mode. On a separate note since you have installed MAS on GCP, my cluster is backed by balanced-pd's, How is the performance overall, any performance related issues?

Also my DB resides outside the cluster and plan to use balanced-pd's for the same. Not sure if they are good enough for a Maximo DB (these are supposedly backed by SSD's)

Again apologies for asking too many questions

Abhinav

Hi Abhinav,

You're right, the choice of ReadWriteMany (RWX) forces you to use such an overkill as filestore. We faced the same challenge and did not want to go this route. What we did instead - switched to ReadWriteOnce (RWO) access mode during the MAS installation:

mas install ... --storage-accessmode 'ReadWriteOnce'

Standard persistent disks in GCP support RWO mode, the PVC will bind to an RWO persistent disk without filestore. While MAS installer pipelines default to RWX for config-pvc in the mas-{instanceid}-pipelines namespace, this runs once and typically doesn't require multi-node access.

Hi All,

I am trying to install MAS on a GCP cluster (3 master 3 worker), at the very end of the installer when it is trying to create the installer pipeline, it is creating a PVC named config-pvc under the mas-{instanceid}-piplines and requesting 500MB worth RWX capable storage. Thing is the only option is GCP capable of RWX is filestore and min 1TB needs to be provisioned which seems unnecessary given these pipelines only need to be run once. Any workaround for this issue?

Abhinav