Question RE Cluster Size during install
RobertH
27 Posts
September 28, 2020, 2:16 pmQuote from RobertH on September 28, 2020, 2:16 pmHave been doing some testing in the lab and are going to be building a semi-live trial run setup (non-mission critical VMs) running off of our live Hyper-V cluster.
During the trial run it will consist of 3 nodes all handling storage with 2x 2.8TB disks in each node plus an NVME for Journal, each node can handle a total of 8 disks.
Ultimately we want to add disks to the nodes as needed (growth wise) and then potentially when the nodes are full add additional nodes to the cluster.
That being said, Im not sure what to choose for the cluster size during setup, do we pick what we will ultimately have disk wise (ie 8*3 = 24ish), or should we choose the smaller "3 to 15" disk option to match the 6 disks we are starting with?
If we choose a smaller quantity to match what we are starting off with, going forward do we just add disks to PetaSAN and it adjusts under the hood or do we need to do something to the pools / ceph to adjust that later on? And if so what needs to be done?
If we have to choose the quantity we will likely end up having ultimately, what happens if we end up exceeding that, for example if we ultimately plan on having 3 total nodes with 8 drives each (24 disks total), and end up adding more nodes down the road?
Thanks for the input
(PS Questions like this would be served well with a FAQ in the quickstart guide, or a pop-up help on the setup wizard)
Have been doing some testing in the lab and are going to be building a semi-live trial run setup (non-mission critical VMs) running off of our live Hyper-V cluster.
During the trial run it will consist of 3 nodes all handling storage with 2x 2.8TB disks in each node plus an NVME for Journal, each node can handle a total of 8 disks.
Ultimately we want to add disks to the nodes as needed (growth wise) and then potentially when the nodes are full add additional nodes to the cluster.
That being said, Im not sure what to choose for the cluster size during setup, do we pick what we will ultimately have disk wise (ie 8*3 = 24ish), or should we choose the smaller "3 to 15" disk option to match the 6 disks we are starting with?
If we choose a smaller quantity to match what we are starting off with, going forward do we just add disks to PetaSAN and it adjusts under the hood or do we need to do something to the pools / ceph to adjust that later on? And if so what needs to be done?
If we have to choose the quantity we will likely end up having ultimately, what happens if we end up exceeding that, for example if we ultimately plan on having 3 total nodes with 8 drives each (24 disks total), and end up adding more nodes down the road?
Thanks for the input
(PS Questions like this would be served well with a FAQ in the quickstart guide, or a pop-up help on the setup wizard)
admin
2,930 Posts
September 28, 2020, 3:14 pmQuote from admin on September 28, 2020, 3:14 pmThis applies to settings the number of PGs for the default pool that gets created on initial cluster deployment.
When you add more disks, the number of PGs in the pool will need be increased at some point, for example if you triple the number of disks. Increasing the number of PGs can be done manually via Ceph commands, it does however rebalance/re-shuffle the stored data so it is a major thing, this is why we do not allow it in the ui so you need to know the impact.
Of course if this is a test cluster, you have the choice to delete data and start a new pool, or you can created a new separate pool when you add disks. Else you will either have to rebalance data. The way to avoid it is to start with the smallest disk count within the selection range.
This applies to settings the number of PGs for the default pool that gets created on initial cluster deployment.
When you add more disks, the number of PGs in the pool will need be increased at some point, for example if you triple the number of disks. Increasing the number of PGs can be done manually via Ceph commands, it does however rebalance/re-shuffle the stored data so it is a major thing, this is why we do not allow it in the ui so you need to know the impact.
Of course if this is a test cluster, you have the choice to delete data and start a new pool, or you can created a new separate pool when you add disks. Else you will either have to rebalance data. The way to avoid it is to start with the smallest disk count within the selection range.
Question RE Cluster Size during install
RobertH
27 Posts
Quote from RobertH on September 28, 2020, 2:16 pmHave been doing some testing in the lab and are going to be building a semi-live trial run setup (non-mission critical VMs) running off of our live Hyper-V cluster.
During the trial run it will consist of 3 nodes all handling storage with 2x 2.8TB disks in each node plus an NVME for Journal, each node can handle a total of 8 disks.
Ultimately we want to add disks to the nodes as needed (growth wise) and then potentially when the nodes are full add additional nodes to the cluster.
That being said, Im not sure what to choose for the cluster size during setup, do we pick what we will ultimately have disk wise (ie 8*3 = 24ish), or should we choose the smaller "3 to 15" disk option to match the 6 disks we are starting with?
If we choose a smaller quantity to match what we are starting off with, going forward do we just add disks to PetaSAN and it adjusts under the hood or do we need to do something to the pools / ceph to adjust that later on? And if so what needs to be done?
If we have to choose the quantity we will likely end up having ultimately, what happens if we end up exceeding that, for example if we ultimately plan on having 3 total nodes with 8 drives each (24 disks total), and end up adding more nodes down the road?
Thanks for the input
(PS Questions like this would be served well with a FAQ in the quickstart guide, or a pop-up help on the setup wizard)
Have been doing some testing in the lab and are going to be building a semi-live trial run setup (non-mission critical VMs) running off of our live Hyper-V cluster.
During the trial run it will consist of 3 nodes all handling storage with 2x 2.8TB disks in each node plus an NVME for Journal, each node can handle a total of 8 disks.
Ultimately we want to add disks to the nodes as needed (growth wise) and then potentially when the nodes are full add additional nodes to the cluster.
That being said, Im not sure what to choose for the cluster size during setup, do we pick what we will ultimately have disk wise (ie 8*3 = 24ish), or should we choose the smaller "3 to 15" disk option to match the 6 disks we are starting with?
If we choose a smaller quantity to match what we are starting off with, going forward do we just add disks to PetaSAN and it adjusts under the hood or do we need to do something to the pools / ceph to adjust that later on? And if so what needs to be done?
If we have to choose the quantity we will likely end up having ultimately, what happens if we end up exceeding that, for example if we ultimately plan on having 3 total nodes with 8 drives each (24 disks total), and end up adding more nodes down the road?
Thanks for the input
(PS Questions like this would be served well with a FAQ in the quickstart guide, or a pop-up help on the setup wizard)
admin
2,930 Posts
Quote from admin on September 28, 2020, 3:14 pmThis applies to settings the number of PGs for the default pool that gets created on initial cluster deployment.
When you add more disks, the number of PGs in the pool will need be increased at some point, for example if you triple the number of disks. Increasing the number of PGs can be done manually via Ceph commands, it does however rebalance/re-shuffle the stored data so it is a major thing, this is why we do not allow it in the ui so you need to know the impact.
Of course if this is a test cluster, you have the choice to delete data and start a new pool, or you can created a new separate pool when you add disks. Else you will either have to rebalance data. The way to avoid it is to start with the smallest disk count within the selection range.
This applies to settings the number of PGs for the default pool that gets created on initial cluster deployment.
When you add more disks, the number of PGs in the pool will need be increased at some point, for example if you triple the number of disks. Increasing the number of PGs can be done manually via Ceph commands, it does however rebalance/re-shuffle the stored data so it is a major thing, this is why we do not allow it in the ui so you need to know the impact.
Of course if this is a test cluster, you have the choice to delete data and start a new pool, or you can created a new separate pool when you add disks. Else you will either have to rebalance data. The way to avoid it is to start with the smallest disk count within the selection range.