[Gluster-users] newbie quick question

Fri Apr 3 20:21:26 UTC 2015

Thanks Joe,  I think that gives me enough to start with now.  I think I'll do 3 x RAID 0 on each server for 3 bricks on each server to start with and see how I get on with that then.

I appreciate the advice, have a good weekend.

Bill

From: Joe Julian [mailto:joe at julianfamily.org]
Sent: 03 April 2015 13:15
To: Bill Dossett; gluster-users at gluster.org
Subject: Re: [Gluster-users] newbie quick question

On 04/03/2015 11:45 AM, Bill Dossett wrote:
Thanks for your response.  My actual need here is to familiarize myself with Gluster...

This is really a lab project for me to try and get as much experience with GlusterFS as I can to see what use it might be.

I've done some thinking and I am going to use the two hosts with 146GB disks in for my oVirt cluster now.

So I am using the 2 x Dell 2950s with 6 x 300GB disks in each for my Gluster.

So my needs now would be that I want as much storage - but able to lose a disk and not lose data, so not RAID 0.

So could you tell me if there are advantages or disadvantages to creating a RAID5 out of the 6 x 300 GB disks on each 2950 server and then allocate whatever is left after loading the operating system and Gluster as a brick?  Or not create RAID and use each disk as brick.  I expect creating the RAID as the Dell Perc controller is pretty efficient would be better than software defined RAID - for performance anyway?

Replication adds fault tolerance, so with replication, raid is often overkill. Use the standard reliability calculations<http://www.eventhelix.com/realtimemantra/faulthandling/system_reliability_availability.htm> to determine if replication alone will meet your SLA/OLA requirements (when/if you have some). The nice thing about replication vs raid is that with replication you can lose virtually (or literally if designed that way) any piece of hardware and still have access to your data.

You may wish to do some raid 0 or raid 1+0 to get enough disk throughput to keep up with your network, depending on each, of course.

Personally, I never do raid 5. The reliability cost vs the cost of storage is just too high for my tastes even if the performance hit wasn't important to my use case.

Thanks again, I really appreciate the advice.

Bill

From: gluster-users-bounces at gluster.org<mailto:gluster-users-bounces at gluster.org> [mailto:gluster-users-bounces at gluster.org] On Behalf Of Joe Julian
Sent: 03 April 2015 12:30
To: gluster-users at gluster.org<mailto:gluster-users at gluster.org>
Subject: Re: [Gluster-users] newbie quick question

imho, start by defining your needs then design the system to meet them. The only goal set out below is that you, "want to get as much storage out of these as possible." This would mean raid0 or 1 brick per disk with a distribute array (no replication).

Identically sized bricks are best. Often if the disk sizes are mixed, raid or partitioning are used to create bricks of the same size.
On 04/03/2015 09:36 AM, Bill Dossett wrote:
Hi, this is my first attempt at Gluster storage.

Just a quick help me get started question - the bare metal page yields a "not found" on the web site so...

I have two dell 2950s fully populated with 5 x 300GB 10K disks.

I want to get as much storage out of these as possible.

I have two more dell 2950s each with 4 x 146GB 10K disks.

Should I build one RAID 5 container on the first two servers, load up CentOS7 and then Gluster on that?  using the unused partition of what is left after loading CentOS?

Or should I be looking at moving physical disks around and putting the smaller 146 GB disks as RAID 1 in each of the 4 servers for OS and distributing the 300GB disks amoung all the servers and maybe pick some more up on Ebay to get all 4 servers identical?

Would appreciate any advice on short notice as I have nothing else to do today for once and would just as soon at least start off on the right foot!

Bill Dossett
Systems Architect

Tech Central - Global Engineering Services

T +1 303 440 3523
M +44 (0)777 590 8612
bill.dossett at pb.com<mailto:bill.dossett at pb.com>
pitneybowes.com<http://pitneybowes.com/>

Pitney Bowes
4750 Walnut Street | Boulder, Colorado, 80301 | USA

In Engineering?
Raise a ticket via Remedy Anywhere [HERE<applewebdata://5D25CA10-0BA9-4233-9DDF-69951F87F8AB/teamsite.inside.pb.com/sites/TCPPM/GES/Lists/Create%20Incident/newIncident.aspx>] takes less than a minute

CloudForms User Guide available [HERE<https://dl.dropboxusercontent.com/u/6219441/ManageIQ%20-%20User%20Manual%20PB%20v5.pdf>]

________________________________

_______________________________________________

Gluster-users mailing list

Gluster-users at gluster.org<mailto:Gluster-users at gluster.org>

http://www.gluster.org/mailman/listinfo/gluster-users

________________________________

________________________________

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20150403/4a36f75d/attachment.html>