VMware Cloud Community
SoheilSP
Contributor
Contributor
Jump to solution

The best way to remove a disk group and add another disk group to vSAN simultaneously

Hi guys,

I have a hybrid vSAN cluster with 13 nodes. Each node has five disk groups, and each disk group has three capacity disks (2TB) and one cache tire. I'm going to remove all capacity disks from disk groups and replace them with one capacity disk (6 TB).What is the best and fastest way to do it? For example, I am going to remove a disk group from a vSAN node with full data migration and add another disk group including one capacity disk (6TB) and one cache disk from another vSAN node in a different fault domain at the same time. Is it a good way to replace it?

0 Kudos
1 Solution

Accepted Solutions
TheBobkin
Champion
Champion
Jump to solution

@SoheilSP, if you want less active time watching/doing it (and have plenty of free space) you could just MM with FDM (Full Data Migration) option one node at a time, remove and recreate all DGs on it once in MM, take it out of MM and move to next node - this might actually be faster than evacuating and removing one DG at a time as there is no chance of data moving multiple times (e.g. from DG1 to DG2, then DG2 to DG3 etc.).

 

Note as this is Hybrid, if you have available slots in the servers for the new disks, you wouldn't even have to remove the DGs - you could just add all the new 6TB disks to each of the DGs then do remove with FDM of all of the existing capacity-tier disks.

 

"I am going to remove a disk group from a vSAN node with full data migration and add another disk group including one capacity disk (6TB) and one cache disk from another vSAN node in a different fault domain at the same time."
I don't understand 2 points here - 1. why would you remove DG on one node and recreate it on a different one? 2. You mention 'Fault Domains'(FDs) - do you have custom FDs configured here or is it stretched cluster or just plain no custom FDs with each node not in an FD?

 

As an aside - why are you replacing 3x disks with 1x disk? You do understand that you will likely have possibly significantly worse performance with this layout compared to the one you currently have?

View solution in original post

0 Kudos
5 Replies
TheBobkin
Champion
Champion
Jump to solution

@SoheilSP, if you want less active time watching/doing it (and have plenty of free space) you could just MM with FDM (Full Data Migration) option one node at a time, remove and recreate all DGs on it once in MM, take it out of MM and move to next node - this might actually be faster than evacuating and removing one DG at a time as there is no chance of data moving multiple times (e.g. from DG1 to DG2, then DG2 to DG3 etc.).

 

Note as this is Hybrid, if you have available slots in the servers for the new disks, you wouldn't even have to remove the DGs - you could just add all the new 6TB disks to each of the DGs then do remove with FDM of all of the existing capacity-tier disks.

 

"I am going to remove a disk group from a vSAN node with full data migration and add another disk group including one capacity disk (6TB) and one cache disk from another vSAN node in a different fault domain at the same time."
I don't understand 2 points here - 1. why would you remove DG on one node and recreate it on a different one? 2. You mention 'Fault Domains'(FDs) - do you have custom FDs configured here or is it stretched cluster or just plain no custom FDs with each node not in an FD?

 

As an aside - why are you replacing 3x disks with 1x disk? You do understand that you will likely have possibly significantly worse performance with this layout compared to the one you currently have?

0 Kudos
NateNateNAte
Hot Shot
Hot Shot
Jump to solution

I was wondering the same thing - why introduce (potentially) more issues for performance or recovery.  I mean, if you're going to replace the equivalent disk capacity with a larger capacity drive - why not introduce 3x 6TB disks then?  

Curious.

0 Kudos
SoheilSP
Contributor
Contributor
Jump to solution

@TheBobkin 

Hi,

We have 13 nodes in our vSAN cluster, and they are in the six fault domains. Once, I tried to remove one 2TB disk from it and add a new 6TB disk to the DG, but after a while, I faced an error, which was cache congestion in that DG, and because of that, I wanted to change my approach. I changed my approach to umount one DG with full data migration and add another DG to another node in a different fault domain, and it seems I've not had any problems yet.

As for your question, we don't have enough space in our datacenter to add another vSAN node to our cluster, and our capacity in our hybrid cluster is going to run out of space. So we decided to change old disks with higher capacity disks which are newer than that. Because of this change I wanted to know what the best way to do it is? We don't have enough 6TB disks to add each DG two disk, after we replace all 2TB disks with 6TB disks to avoid reducing performance we will be adding another 6TB disk to each DG.

0 Kudos
SoheilSP
Contributor
Contributor
Jump to solution

@NateNateNAte 

Hi,

We don't have enough 6TB disks and we ordered extra disks! 🙂

After we replace all 2TB disks with 6TB disks and we get extra disks, we'll intent on adding another disk to each DGs.

0 Kudos
NateNateNAte
Hot Shot
Hot Shot
Jump to solution

@SoheilSP ,

That makes sense then. If the 2Tb disks are still viable, and understanding you have limited space in your server room, would you keep those as spares, or populate a spare appliance so that you have a stand-by or back-up storage volume available? Just curious.  I understand the approach your taking and that's a reasonable way to make that type of upgrade.  I hope it continues to operate well for you.

0 Kudos