Expend CEPH pool storage

[lstrunjak]

Hello,

I need help or advice. We have in our environment 6 servers. Every server have 2 disk 4TB each disk for CEPH storage. Se we have 12 hdd/osd in CEPH pool. So now we add 2 more disk in every server and want to expend CEPH pool. How to do that? is there any chance to crash existing pool by adding new disks? I've attached screen shot of CEPH configuration. Proxmox is: 6.3-4 and ceph is: 15.2.8.


Please help me

 

[spirit]

you just need to add osd, a pool always use all disk capacity.

(be carefull that datas are auto rebalanced to the new disks, so expect to have network traffic && disk access)

 

[lstrunjak]

So I do not need to change any PGs or anything. It will be done automatic. So I just need to add new disk/osd to the existing pool? And data and everything will be safe

 

[lstrunjak]

Please anyone? related to the configuration can I just resize pool?

 

[RokaKen]

Please anyone? related to the configuration can I just resize pool?

Technically, "resize" refers to the number of replicas, so no, don't change size. However, you are asking about PG count -- yes, I would increase PG count to at least 1024 for a total of 24 OSDs. If you use the PG autoscaler [0] (default enabled), it will change the PG count automatically.

[0] https://docs.ceph.com/en/octopus/rados/operations/placement-groups/#

 

[aaron]

So I just need to add new disk/osd to the existing pool? And data and everything will be safe

Data should always be safe. The number of PGs is used to determine in how many placement groups the objects should be grouped into.
The optimal number of PGs depends on the number of pools, their size and the number of OSDs, as each OSD should have around 100PGs.

You can check with ceph osd df tree how many PGs are stored on your OSDs.
The autoscaler seems to be on. It will become active once the optimal number of PGs is different to the current PG number by a factor of 3. Before that, you will have to set the PG number manually.
For the autoscaler to work best, you should tell it where the journey is going with the target_ratio / target_size parameter of the pool. In your case it is easy and I would set the target_ratio to 1 for the pool CEPH. The device health metrics pool is there for maintenance reasons and will only consume very little space and can be ignored for this.

You can also use the PG Calculator to see what would be ideal in your situation. If i understand you correctly, you will have 24OSDs in the end?