HPE Storage Users Group

A Storage Administrator Community




Post new topic Reply to topic  [ 5 posts ] 
Author Message
 Post subject: Testing Thinly-Deduped volumes in new 3PAR and CPG doubt
PostPosted: Tue Jul 17, 2018 9:38 am 

Joined: Mon Apr 09, 2018 7:19 am
Posts: 13
Hello again!!

I'm running my first tests in our new 8440 array. I'm new to the deduplication feature (system is at 3.2.2 firmware so no compression yet).

I've created a LUN to present it to a Windows physical server and try to move data on it. All Ok, deduplications savings say near 0 but is normal because I copied in a veeam backup that is already deduped and compressed.
Later I tried some user files...very few data but ok.

The "problem" is that I have deleted the VV but I continue to see almost 100GB of occupation in the CPG (all data I moved for testing). So it's like is not recovering space after volumes deletion.

Is this behaviour related to deduplication in any way?? What I have to do to have all the space back to free?? I asume deleting the CPG will clear all data but it should exist another way, isn't it??

Also another question... which RAID level and set size for 20 x 1,92TB SSD??
HP says in his Best Practices document:
Solid-state drive CPGs
Best practice: Solid-state drive (SSD) CPGs should be of the RAID 6 type with the default set size. This will bring superior performance/capacity ratio and 7+1 on systems that have enough SSDs to support 7+1. If maximum performance is required, use RAID 1.

So this information is incorrect because a 7+1 is a RAID 5, not a RAID 6. Which part is incorrect?? the RAID 6 or the 7+1??

The technician who installed the array told me that RAID 5 was fine and that I could make the set size bigger than default.
For me RAID 5 (7+1) sounds well. Is this a good configuration for 20 disks??

BR.

EDIT: LOL! I have created and removed two more LUNs, and launched a compact CPG in the console and now says 0% occupation, so it seems I only need recommendations for the second question. :)


Top
 Profile  
Reply with quote  
 Post subject: Re: Testing Thinly-Deduped volumes in new 3PAR and CPG doubt
PostPosted: Tue Jul 17, 2018 10:50 am 

Joined: Wed Nov 09, 2011 12:01 pm
Posts: 392
Yes typically a compact will free up space but deleting all the VV tends to work too.

I think HPE is now pushing RAID6 as preferred for most things, it used to be for all NL and any SSD >2TB but I've now seen it recommended for FC and smaller SSDs.

It's to do with rebuild times and improving resilience with larger drive sizes, most vendors seem to be moving to a multi-parity model for enterprise storage now.

I've also seen the confusing use of x+1 and RAID6, basically just double the numbers, e.g. 7+1 -> 14+2.

As to correct number, it depends on your system setup, based on the number of drives attached to each Node Pair. If your 8440 has 4 Nodes and 20 disks then you should have 10 disks per Node Pair, the disks are then halved for per Node so 5 disks so I think default of 4+1 (8+2). They did add a feature that bypasses the per Node limit so you maybe able to push that higher but not beyond the Node Pair max.

Also beware that CPG might not grow if you have a number of disks failed preventing it allocating the stripe size so avoid maxing the size.

You might also want to look at upgrading to 3.3.1MU2 before using dedupe, otherwise you'll need to find the space to undedupe everything then rededupe if upgrading later.


Top
 Profile  
Reply with quote  
 Post subject: Re: Testing Thinly-Deduped volumes in new 3PAR and CPG doubt
PostPosted: Thu Jul 19, 2018 2:12 am 

Joined: Mon Apr 09, 2018 7:19 am
Posts: 13
ailean wrote:
I think HPE is now pushing RAID6 as preferred for most things, it used to be for all NL and any SSD >2TB but I've now seen it recommended for FC and smaller SSDs.

As to correct number, it depends on your system setup, based on the number of drives attached to each Node Pair. If your 8440 has 4 Nodes and 20 disks then you should have 10 disks per Node Pair, the disks are then halved for per Node so 5 disks so I think default of 4+1 (8+2). They did add a feature that bypasses the per Node limit so you maybe able to push that higher but not beyond the Node Pair max.

Also beware that CPG might not grow if you have a number of disks failed preventing it allocating the stripe size so avoid maxing the size.

You might also want to look at upgrading to 3.3.1MU2 before using dedupe, otherwise you'll need to find the space to undedupe everything then rededupe if upgrading later.


Hello, Ailean :)

My 8440 has 2 nodes, so 1 node pair with 20 disks SSD 1,92TB

So if I've understood correctly, my ideal setup is RAID 6 (8+2), so 10 disk each node.
But what happens if I loose a disk? It simply writes the missing chunklet in a disk of the other node?? And if I add more disks in a future?? I need to rebuild CFGs with new structure?? for example, if I add 4 disks, move to 10+2 set size??

I was planning to do a RAID5 7+1 setup but I think RAID6 8+2 would be also ok. A bit less available space but enough... or leaving the default RAID5 5+1

Which option would you choose???


I'm waiting to have the firmware upgrade before doing anything in the array.

BR.


Top
 Profile  
Reply with quote  
 Post subject: Re: Testing Thinly-Deduped volumes in new 3PAR and CPG doubt
PostPosted: Thu Jul 19, 2018 2:43 am 

Joined: Mon Sep 21, 2015 2:11 pm
Posts: 1570
Location: Europe
Dave338 wrote:
My 8440 has 2 nodes, so 1 node pair with 20 disks SSD 1,92TB

So if I've understood correctly, my ideal setup is RAID 6 (8+2), so 10 disk each node.
But what happens if I loose a disk? It simply writes the missing chunklet in a disk of the other node??

Remember that the 3PAR "doesn't care" about disks when allocating space, only chunklets. What will happen is that when it can't create a stripe on the local node, it will switch to express layout and take chunklets from SSDs previosuly only used by the other node.
Dave338 wrote:
And if I add more disks in a future?? I need to rebuild CFGs with new structure?? for example, if I add 4 disks, move to 10+2 set size??

You could. Simply change existing CPG and run tunesys and you changed that for all existing and new data. But you don't have to....
Dave338 wrote:

I was planning to do a RAID5 7+1 setup but I think RAID6 8+2 would be also ok. A bit less available space but enough... or leaving the default RAID5 5+1

Which option would you choose???

No vendors are recommending single parity. Bigger drives and larger sets is increasing the risk of an URE during rebuild. https://h20195.www2.hpe.com/v2/getdocum ... 0000244enw, page 7.
Dave338 wrote:

I'm waiting to have the firmware upgrade before doing anything in the array.


Smart choice. Not only will it give you compression, but it will also give you a new dedupe version (assuming you're starting fresh with a new CPG).

_________________
The views and opinions expressed are my own and do not necessarily reflect those of my current or previous employers.


Top
 Profile  
Reply with quote  
 Post subject: Re: Testing Thinly-Deduped volumes in new 3PAR and CPG doubt
PostPosted: Thu Jul 19, 2018 3:52 am 

Joined: Mon Apr 09, 2018 7:19 am
Posts: 13
MammaGutt wrote:
Remember that the 3PAR "doesn't care" about disks when allocating space, only chunklets. What will happen is that when it can't create a stripe on the local node, it will switch to express layout and take chunklets from SSDs previosuly only used by the other node.


I know, I know.. I have another 3PAR (7200c) in another site, but when it was initialized was my first 3PAR and i let the provider's expert take the decisions. Now is my turn :)
That "express layout" thing, is bad?? When I replace the faulty disk, all goes back to "normal" layout automatically?? I have to do a tunesys, tuneld... etc??

MammaGutt wrote:
Smart choice. Not only will it give you compression, but it will also give you a new dedupe version (assuming you're starting fresh with a new CPG).

Yes, I will create new CFGs after the upgrade and start working in data migration from our old EVA6550 (I'll do it at host level, moving 1 FC from each host to the new FC switches and then storage vmotion)

Thanks and BR :)


Top
 Profile  
Reply with quote  
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 5 posts ] 


Who is online

Users browsing this forum: Google [Bot] and 44 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group | DVGFX2 by: Matt