This article will walk through the process of creating an Erasure Code Pool in PetaSAN along with creating the rules you will need for said pool
A minimum of 3 nodes active and running
- The First step is creating a rule for the type of redundancy you want for the pool and the failure domain. Go to wrench icon –>CRUSH –> Rules. For our example we are going to choose our failure domain at the host level, and it will have erasure code redundancy. we want this pool to only utilize the ssd’s so our option is ec-by-host-ssd. press save.
- You must also create an EC Profile to do this go to the wrench icon –> EC Profiles. You’ll give it a name select your K amount and your M amount we’ll be doing a 2+1 Erasure code so K=2 M=1 and press save. your screen will show your newly created profile here:
- At this point we are good to go ahead and create our pool. Go to the wrench icon –> pools–> add pool. At this point we will name our pool, we’re going to select “EC” for this example we want to pool to serve out iSCSI disks so for usage we will select “rbd” . We will then select out EC Profile since in this example it is a cluster with three management nodes and one storage node we will select ” ec-21-profile” which will give us our 2+1 erasure code. We then select our number of PGs. The PG number is reliant on the size of your cluster in regards to number of OSD’s and the density of those OSD’s, this can be calculated by using the tool PGcalc listed here: https://ceph.io/pgcalc/ we will choose 64 since this is a very small cluster. The Size is automatically taken care of, which brings us to min size, (This should be greater than or equal to K and less than or equal to K+M) since we set up a 2+1 EC the default is set to 3, we won’t worry about compression for now. Finally we must select a rule name, since we created it earlier it’s just a simple matter of selecting “ec-by-host-ssd” and save.
- Once the pool successfully creates you will have a screen similar to the following:
- If you receive the previous post “Pool created successfully)
- you can also ssh into a PetaSAN node and run ” ceph osd pool stats”
- If you do not see an option for an EC Profile or Rule Name ensure they were saved.