ZFS pool configuration same name

[halley80]

Hello,

When creating a zfs pool (raidz) on the command line, it does not appear in the dashboard.
If I create it through the graphical user interface, I get an error message informing me that the pool already exists on the other node.
Cluster:
2 nodes pve 5.3-8 / local-zfs+datastore raidz

thank you in advance for your help

 

[mira]

If you create the pool via CLI you'll have to add its storage definition in Datacenter -> Storage. If you do it via the GUI it's automatically added as long as the 'Add Storage' checkbox is set.

 

[halley80]

Hello,

Thank you for your answer.
My goal is to be able to migrate vm from a ZFS datastore node 1 <=> node 2.

It works perfectly with the local ZFS pool but not with my datastores that have a different name.

I seem to remember that during a previous installation:
When I created the datastores (ZFS raidz), with identical names, before the cluster was created, it worked perfectly.
Having only linux vm, is the solution to use ZFS share in NFS mode ?

Thank a lot for your advice.
kind regards

My pools:
zpool list => node 1
datastore_10k 6.50T 816K 6.50T - 0% 0% 1.00x ONLINE -
rpool 556G 61.8G 494G - 0% 11% 1.00x ONLINE -

zpool list = node 2
datastore_10k_2 6.50T 673G 5.84T - 1% 10% 1.00x ONLINE -
rpool 556G 19.4G 537G - 0% 3% 1.00x ONLINE -

 

[mira]

Please post your storage config (/etc/pve/storage.cfg).

 

[halley80]

My configuration storage file below:
*********************************************
dir: local
path /var/lib/vz
content vztmpl,iso,backup

zfspool: local-zfs
pool rpool/data
content images,rootdir
sparse 1

zfspool: datastore_10k_2
pool datastore_10k_2
content images,rootdir
nodes ipmpve2

nfs: nas_dell
export /stockage/save_ipmprod
path /mnt/pve/nas_dell
server 192.168.38.50
content images,backup
maxfiles 2
options vers=3

zfspool: datastore_10k
pool datastore_10k
content rootdir,images
nodes ipmpve1
sparse 0
*********************************************

 

[mira]

When migrating online you can specify a targetstorage (see 'man qm'), but offline you need the same storage on the target node.
It's possible to rename a zfspool via 'zpool export <pool>' followed by a 'zpool import <pool> <new-pool-name>' (see 'man zpool').
With this you could remove the datastorage_10k_2 entry (or the datastore_10k entry, whichever pool you remove) and have just a single storage available on both nodes.

 

[halley80]

Hello,

Thank you very much for your answers.
For simplicity, a screenshot of the cluster as an attachment (I am in production)
I quickly plan a pre-production cluster

1/ So for an online migration in my case:
vm-id-100=> node2 :
"qm migrate 100 ipmpve2 --targetstorage datastore_10k_2 --online --with-local-disks" ?

2/ To make the ZFS datastore_10k pool available on all node :
On node 1 :
"zpool export datastore_10k"
On node 1/2 :
"zpool import datastore_10k datastore_10k_1" ?
I would then see the datastore_10k_1 on the 2 nodes ?
(Sorry, I read the man pages but if it's not yet clear to me)

3/ last two administrative questions:
We will subscribe to commercial support.
We are a university (France) and for the necessary documents, I need to know:
In which country are you located ?
What payment method do you accept ?
https://www.univ-littoral.fr/

Thank you again for your advice
kind regards

 

[mira]

1) Yes, that's how you do online migration with local disks. Currently it's only possible via CLI.
2) It's the other way around. If you want to rename your 'datastore_10k_2' to 'datastore_10k' so you have on both nodes the same name, you export 'datastore_10k_2' and import it as 'datastore_10k':

zpool export datastore_10k_2
zpool import datastore_10k_2 datastore_10k

If you want to rename 'datastore_10k' to 'datastore_10k_2' then simply swap those 2 in the above commands.

3) Take a look here: https://www.proxmox.com/en/legal-notice
See also: https://www.proxmox.com/en/proxmox-ve/pricing and the subscription agreement: https://www.proxmox.com/downloads/item/proxmox-ve-subscription-agreement
If you buy via https://shop.maurer-it.com/ you can find all the information here: https://www.maurer-it.com/en/about/faq
You can also take a look at our partners, perhaps one of them fits your needs better: https://www.proxmox.com/en/partners

 

[halley80]

thank you again for all your answers !

1/ I now understand, however the renaming of the ZFS pool "datastore_10_2" implies the migration of vm(s) to the "datastore_10K" ?
is there a risk of data loss or something else (cluster problem) after the "datastore_10K_2" renaming "datastore_10K" ?
I'm thinking of the".conf" files of the vm(s) in /etc/pve/nodes/ipmpve2/qemu-server
=> tag scsi0 and vmgenid

2/ I have read many posts to improve the performance of ZFS pools
it is recommended to add cache via SSD disk(s)

Specs of my cluster:
2 Dell R440 - 2 x Xeon Bronze 8 cores@1.7 Ghz - 64 GB RAM
two ZFS pools per server :
system => ZFS mirror 10K (2x600Gb)
data => ZFS RAIDZ-1 10K (4x1.8TB)

which SSD disk size is required?
I searched on the internet if there is a calculation method for the cache without success

kind regards

 

[mira]

1) You'll have to change the config entries of every VM with a disk on the pool.

2) For ZIL/SLOG here's a nice presentation with a lot of information: http://open-zfs.org/w/images/c/c8/10-ZIL_performance.pdf and another article https://www.servethehome.com/what-is-the-zfs-zil-slog-and-what-makes-a-good-one/
For SLOG size there's also this: https://www.ixsystems.com/blog/why-zil-size-matters-or-doesnt/
For Cache it's not that easy. Bigger 'l2arc' takes up more space in 'arc'. So probably not the best solution with as little as 64GB RAM if you're also running VMs on the machine. The cache is just a read cache. You could try your setup first and if reads are too slow try adding a cache device (can be added and removed easily).

 

[halley80]

Thank you again for everything.
it works perfectly

kind regards