VMware Cloud Community
Lalegre
Virtuoso
Virtuoso

VSAN 6.6.1 - Erasure Coding - Deduplication and Compression

Hello folks,

I am having this issue where i converted a lot of Virtual machines from RAID 1 vSAN Policy to RAID 5 Erasure Coding. I did this to gain more space on my vSAN Datastore but i just only gain about 800 GB and converted 50 VMs consuming aroung 10 TB totally.


This picture is before applying Erasure Coding

pastedImage_0.png

This picture is after applying Erasure Coding

pastedImage_1.png

Is this a normal behaviour or is an expected issue?

0 Kudos
4 Replies
Lalegre
Virtuoso
Virtuoso

Any idea on this topic?

Thanks!

0 Kudos
TheBobkin
Champion
Champion

Hello Lalegre​,

Response to PM (sharing here in case it helps anyone else understand what to look at/for):

"And what if i don't have any of both scenarios, if i check with RVC i don't have any "proportionalCapacity" unless i specify it on the Policy."

If it has no proportionalCapacity parameter defined it should be thin - you should be able to easily check this by comparing the usage of the data components and the (max ever used IIRC) in-guest disk usage e.g. if I make a Thick 1TB FTT=1 FTM=RAID1 vmdk it will be composed of  5x R0 200GB components per R1 mirror, but if I make this Thin each component would be tiny (until you add data to it); you can compare this for some Objects with each FTM applied - FTT=1 FTM=RAID1 OSR=0 should be ~200% the max used disk space as presented to the guest-OS, FTT=1 FTM=RAID5 OSR=0 should be ~133%.

So your screenshot looks to have something up with it e.g. before with FTT=1 FTM=RAID1 is 36TB used so 18TB by 1.33 is 24TB not 33.6TB - are you 100% not-a-shadow-of-a-doubt that no data was FTT=0 before making them FTT=1 RAID5? If you have output of python /usr/lib/vmware/vsan/bin/vsan-health-status.pyc from before and after then should be able to determine this.

Anyway, so less dedupe ratio aswell could be from parity data not deduping as well (not positive) or just the new placement of the data on the disk-groups has less commonality than it had.

"Also the total space consumed by the VMs is 8 TB"

What do you mean here specifically? (e.g how are you calculating this and where)

Bob

0 Kudos
Lalegre
Virtuoso
Virtuoso

Let me explain,

I have an environment where i have a lot of VMs and converted from RAID 1 to RAID 5, 8 TB of VMs provisioned with Thin Provisioning.

After the conversion, the dedupe ratio decreased 0.10 x and the "USED BEFORE" of the Deduplication and Compression Overview decreased only 2 TB.

So i don't understand why having 8 TB RAW occupied (Using RAID1 16TB) only decrease 2 TB if i apply RAID5 and why the deduplication ratio decreases

0 Kudos
TheBobkin
Champion
Champion

Hello Lalegre​,

"So i don't understand why having 8 TB RAW occupied (Using RAID1 16TB) only decrease 2 TB if i apply RAID5 and why the deduplication ratio decreases"

Where/how are you measuring "8TB RAW" of data?

Okay, so you have only changed some of the data from RAID1 to RAID5 - I would advise you look closer at the actual structure and usage of the components/Objects before and after changing the Storage Policy to verify how much they are using before and after.

"and why the deduplication ratio decreases"

Potentially as I said in the second-last sentence of my previous comment.

Bob

0 Kudos