Question of MapR 6.1 File System

[cloverlu]

We want to live expand mapr-fs.

For example, we have one 70GB fs and another 150 G C fs. The 70GB fs is showing some storage occuplied. however, It doesn't recognize the 150GB additional mapr-fs. Even if the 70GB is full, the 150GB file allocated won’t be utilized. How do we live expand the mapr-fs?

[tdunning]

In order to answer this question, there needs to be a bit more in the way of details. In general, live expansion works very nicely with the HPE Ezmeral Data Fabric (used to be mapr fs). There are several ways to go wrong, however, and even if you do it right, it is very easy to be surprised by the fact that the file system may choose not to rebalance existing data across the newly expanded cluster. If you are adding a server or three to a large very full cluster, you could see an apparent degradation in performance if all new data goes only to the new servers.

In general, however, you can add drives or nodes and expect to see expansion pretty transparently. 

If you need more specific, blow-by-blow advice, there is more information that is really needed.

So, first off, can you say how many nodes you have in the original cluster and how many are in the expanded cluster? How many disks in each?

How did you install the original, and how did you add more storage? Did you add new servers? Or just add drives to existing servers?

Also, did you really mean that you are storing less than 200Gb of data? Are you saying that your entire cluster has less than 200GB of storage? Or do you mean that you are adding to a dataset on a larger cluster?

[darwink]

This is our integration test environment. We have only one node with 70GB as data disk when we first installed MapR 6.1. Now we want to add more data disks to this node, so we attach one new disk of 150GB and add it as part of mapr-fs by MapR admin UI.

What is the correct process to add new data disks to an existing node? Is it related to storage pool allocation? If yes, how to make change?

[AKChoudhury]

What is the correct process to add new data disks to an existing node?

If you would like to add more storage to the existing mapr node, you can add it by creating new storage pools on that specific node.

Please follow the below article for more information.

https://docs.datafabric.hpe.com/62/ClusterAdministration/nodes/Node-AddDisks.html

[Shishir_Prakash]

that's correct, you can add disks on an existing node to expand the cluster storage capacity.  

By default the stripe width of storage pool is 3 which means if you add three disk using below command then it will create 1 SP on that node.

maprcli disk add -disks <disk names> -host <host>

Note: You can create SPs with one or more disks. In general if the disks are spindle disks (HDD) then recomondations are to use minimum three disks per SP. Whereas if using SSDs then you can use 1 disk per SP. 

[Shishir_Prakash]

You can use below doc to add the disk in maprfs.  

https://docs.datafabric.hpe.com/62/ClusterAdministration/nodes/Node-AddDisks.html

The newly added disk will be part of new SP (storage pool). You can not extend an exisiting SP so this newly added disk will be part of new SP. 

You can list the SPs and the disks associated with the SPs using below command.

/opt/mapr/server/mrconfig sp list -v