Jump to content

Recommended Posts

Posted
2 hours ago, PotNoodleTech said:

 

So..out of interest if you had a server fail would you never buy another server again and install everything on the PCs instead?

 

If your car broke down would you never buy another car?

 

Just wondering as raid card failure is pretty darn rare - never had one myself in 24 years of doing this.

 

Actually, since we virtualize, I would simply move the server to another host ;)

 

What's worth mentioning, is that RAID actually doubles or triples the chance of a disk failure straight away by its very design. Also the RAID card, as well as any software issues can also corrupt storage. Particularly if they lose power at boot time or during a rebuild.

 

I mean in the old days of spinning rust and small hard disks, RAID was a useful technology. But these days its been made redundant by newer technologies and recovery strategies. As an example you won't go into a data center and find every server with RAID 5 setups.

 

I've just found over the years that having a single reliable SSD disk with a good backup strategy is a far better choice.

  • Like 1
Posted (edited)
1 hour ago, supportman said:

 

What's worth mentioning, is that RAID actually doubles or triples the chance of a disk failure straight away by its very design. Also the RAID card, as well as any software issues can also corrupt storage. Particularly if they lose power at boot time or during a rebuild.

 

 

Hehe - it's all a game of chance. I don't disagree with you SSDs Virtualisation and Veeam are a godsend for reliability.

 

But the chance of one disk failing is not the same as the chance of your server being down, wich is the actual important thing.

 

If you have one disk the the chance of the disk failing is whatever it is, but if it does fail the chance of your system being down is 100%

 

If you have 5 disks in Raid 5 the chance of one disk failing is 5 x higher than whatever the singe disk chance of failing above, BUT the chance of your system being down due to one disk failing goes down to 0% as the system can tolerate one disk failing and stay up.

 

Therefore I will keep using antiquated RAID as it literally just reduces the chance of your sever going down.

Edited by PotNoodleTech
  • Like 2
Posted
2 minutes ago, PotNoodleTech said:

Therefore I will keep using antiquated RAID as it literally just reduces the chance of your sever going down.

Now waiting for name_redacted to pop up and berate us for having to choose between RAID6  or services that can be taken out by a monolithic disk/server failure, when we should be using kubernetes or something.

  • Haha 2
Posted

Interesting idea, that good quality SSD and rapid recovery would now supersede traditional hardware RAID.

 

For our part, for the first time in a decade, we've just had a disk fail in our main production server, spinning rust from the main RAID5 storage array. I wouldn't have known, if the server management hadn't raised an alert to tell me. New drive is on order for tomorrow. So, hardware RAID working just as it should for us. We only have one server, so for us the redundancy has to be in the disks. 

 

I wonder if we had a cluster of at least two, would we just use high quality SSD and let the whole servers be the redundancy? It's an idea, I suppose. I'd probably still be likely to do hardware RAID with SSD plus redundant servers in an ideal scenario, belt and braces, but that's out of habit. 

  • Like 1
Posted

RAID 6 for everything, double redundancy, yes it is the slowest but I want peace of mind and the performance is a second place issue.  No need to rush the fix. 

 

I hate the RAID 5 re-build or worse wait for a new HDD before the re-build (or new hot spare).. The re-build, the most extensive workout for the array, so the perfect time to find another marginal drive!

 

 

Posted

I'm still shocked that anyone even considers hardware raid in 2025. ZFS has been around for DECADES! Is this just a case of 'i can't find an .exe' for it so won't learn? Or something else?

Posted
On 31/03/2025 at 15:42, psydii said:

Now waiting for name_redacted to pop up and berate us for having to choose between RAID6  or services that can be taken out by a monolithic disk/server failure, when we should be using kubernetes or something.

Oh was that me. Yeah get your sh*t together.

  • Haha 1
Posted
1 minute ago, junz said:

Does Windows support ZFS?

Yes, depending on how you run it. The last place I worked ( in education) we had 4000 windows desktops and all our servers running on ZFS. We could snapshot every 15 mins. 

Posted (edited)
11 hours ago, dmj said:

I'm still shocked that anyone even considers hardware raid in 2025. ZFS has been around for DECADES! Is this just a case of 'i can't find an .exe' for it so won't learn? Or something else?

 

Cool - insult everyone that doesnt use ZFS.

 

Some of us are sitting back chuckling at all the people running Windows 2025 DCs and having their entire network go down due to DHCP bugs etc.

 

Some peeps want to be on the bleeding edge and that's cool, some people know what works well and just sit back and admire their 100% uptime. That's cool too.  There's room for both!!

 

 

Edited by PotNoodleTech
  • Like 2
Posted

In a world of limited budgets, no training and no time to look into anything, a lot of solutions are making the best of the situation you have :-

 

Two hosts running Smart spaces, standalone - 4 HDD and 4 SSD - Worked ok but once it started filling up the lag when it fell back to HDD tier was terrible

 

We went for a HP3PAR All SSD RAID6 - our demand for storage skyrocketed thanks to Videos being used, electronic evidence for OCR as an example can be 25gb per file per Student

 

Testing we went with mixture of Crystal Diskmark and ATTO

  • Like 1
Posted
11 hours ago, dmj said:

I'm still shocked that anyone even considers hardware raid in 2025. ZFS has been around for DECADES! Is this just a case of 'i can't find an .exe' for it so won't learn? Or something else?

ZFS might have been around for a decade, but it's still way less known about than RAID. If you Google "disk redundancy" you'll see a lot of stuff about RAID and virtually none about ZFS. So it's one of those things that even in tech circles people don't know they don't know about it, and it's unlikely to cross their paths organically. 

That and the old "if it's working, why change it?"

 

 

Posted
8 hours ago, PotNoodleTech said:

 

Cool - insult everyone that doesnt use ZFS

 

 

No insult. I appreciate schools are behind on budget, training etc but this one is a no brainer for me - Free software that's been developed since the days of WindowsXP, super stable and in use by many large orgs.

 

 

7 hours ago, Rob_D said:

ZFS might have been around for a decade, but it's still way less known about than RAID. If you Google "disk redundancy" you'll see a lot of stuff about RAID and virtually none about ZFS. So it's one of those things that even in tech circles people don't know they don't know about it, and it's unlikely to cross their paths organically. 

That and the old "if it's working, why change it?"

 

 

I guess that's what edugeek is for, to help understand some of the options. "if it's working, why change it?" a couple of things: RAID6 doesn't do dedupe, replication, snapshot to remote locations,  and isn't suitable as a backup. You could save money on you backup solution, not worry about adding raid cards to your servers: simply it's a better solution more features at a lower cost. 

Posted
3 hours ago, dmj said:

[RAID6] isn't suitable as a backup. You could save money on you backup solution,

Are you saying ZFS can be used as a backup solution?

Posted

My server runs ZFS, it does a snapshot of each VM every 2 hours and sends the delta to the backup server that reintegrates it into its own zfs copy. Backup server keeps many months and years of snapshots, auto deletes the hourly ones after 2 months. If we got ransomewared we'd only need to go back 2 hours.

 

Also the only way to delete the backup is to physically be logged into the system locked in the cabinet, the only thing you can do remotely is send it snapshots

  • Like 1
  • Thanks 2
Posted
3 hours ago, mavhc said:

My server runs ZFS, it does a snapshot of each VM every 2 hours and sends the delta to the backup server that reintegrates it into its own zfs copy. Backup server keeps many months and years of snapshots, auto deletes the hourly ones after 2 months. If we got ransomewared we'd only need to go back 2 hours.

 

Also the only way to delete the backup is to physically be logged into the system locked in the cabinet, the only thing you can do remotely is send it snapshots

 

the University I was working at did essentially the same thing - but 15min snapshots, which was a lifeline for users who needed to revert a document before the days of google docs. 

Posted
3 hours ago, PotNoodleTech said:

So whats the easiest way of getting ZFS running then?

 

zpool create

😜

 

 it rather depends on what you are starting with and what the use case is. I guess QNAP, TrueNAS or similar is the easiest 

  • Like 1
Posted (edited)
17 hours ago, dmj said:

I guess QNAP, TrueNAS or similar is the easiest 

 

52 minutes ago, mavhc said:

Install Proxmox on the bare metal, it'll even boot from a raidz, then run your Windows VMs on that.

 

So convert your infrastructure to linux? That's probably another barrier to entry in a lot of Edu where it's primarily windows domains.

If I were rebuilding the system from the ground up then I'd be up for doing it. But I've only ever come close to that level of rebuild once in 15 years.

Edited by Rob_D
Posted
36 minutes ago, Rob_D said:

 

 

So convert your infrastructure to linux? That's probably another barrier to entry in a lot of Edu where it's primarily windows domains.

If I were rebuilding the system from the ground up then I'd be up for doing it. But I've only ever come close to that level of rebuild once in 15 years.

You wouldn't need to change anything from the user or domain perspective. It's no different than running VMWare etc. It's not a huge change. 

  • Like 1
Posted
5 minutes ago, dmj said:

You wouldn't need to change anything from the user or domain perspective. It's no different than running VMWare etc. It's not a huge change. 

The same could be said for server virtualization, that took us 4-5 months. You're still talking about formatting down and rebuilding the hosts, then migrating all the VMS across.

I'm not saying it's not doable, or event that technically complex, but it's still a pretty big job.  

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...