Koozali.org: home of the SME Server

Obsolete Releases => SME Server 9.x => Topic started by: nicolatiana on July 04, 2019, 06:04:35 PM

Title: Partition scenario/size
Post by: nicolatiana on July 04, 2019, 06:04:35 PM

Let's consider a couple of Hp Microserver Gen10 with 4x4 TB hdd to be configured in RAID 5 no spare to get 12 Tb of available space to backup with Affa.


According to specifications Centos 6.x + mdadm should support such a configuration.


Anyone that has tested it "on the field" ?


Nicola

Title: Re: Partition scenario/size
Post by: janet on July 04, 2019, 06:33:32 PM

nicolatiana

RAID howto implies it should be OK
https://wiki.contribs.org/Raid

Title: Re: Partition scenario/size
Post by: ReetP on July 04, 2019, 07:21:48 PM

Can you? Yes.
Should you? Hmmm. That's the question you should really be asking.

If you are happy taking the risk your backup machine will fail completely whilst rebuilding after a drive failure.

Which will almost certainly occur because your real server has had a catastrophic Raid failure. The law of sod & all that jazz.

Raid 5 with large drives is not the greatest of ideas, and even Raid 6 is flakey at these drive sizes.

Worth some serious thought.

Title: Re: Partition scenario/size
Post by: janet on July 05, 2019, 06:00:04 AM

nicolatiana

Further to Reet

https://www.zdnet.com/article/raidfail-dont-use-raid-5-on-small-arrays/

Title: Re: Partition scenario/size
Post by: nicolatiana on July 05, 2019, 08:07:09 AM

You are right but the main problem is as always "budget".


I know that other solutions would give me more efficiency and safe even if a RAID 6 or 10 too with only 4 drives available does not give more security than RAID 5 but I should try to give the best possible with a 5000/6000 budget.


Two further ways for me, both with some question marks:


1) Using the same solution (a couple of MSGen10) with a Raid1 solution based on 10/12 WDNasWarePro Red disks; the question is: disk so large are supported during install and running ? My experience stops at 4Tb in mirror/R1.


2) More expensive: a couple of Qnaps 6-8 bays (with more raid options) installing Sme inside Qnaps Virtualization Station but I know too few about this (I suppose it's based on KVM/Qemu) and about its performances.


The backup scenario is a 9.2 server (in Esxi 6.5) with about 2,9 Tb data growing in backup up to 5,4 due to usual/physiological file modification and, of course, users devoted to change frequently name to 100 Gb folders  :( [size=78%] .[/size]