Upgrading server with RAID array

[Insomniator]

So I have a Dell server that currently has two SATA drives in RAID 1. I am going to be installing new SSD's in RAID 1 plus a new OS and then copying the data over.

My question is if required, will I be able to just plug back in the old drives if something doesn't work and I need to have the old system back to normal? Or is the old raid array gone once I create a new one with the new drives.

Just looking for what my emergency 'oh ****' plan would be if I take the old drives out, setup the new raid array on the new drives and windows doesn't install or something.

 

[Viper GTS]

You'll need to import the foreign config, but other than that yes if you don't over-write them you'll be fine.

Viper GTS

 

[Oyeve]

You'll need to import the foreign config, but other than that yes if you don't over-write them you'll be fine.

Viper GTS

Would this work in a SW and HW raid?

 

[BonzaiDuck]

RAID 1 provides redundancy, but it isn't "backup." It means that your server would continue to operate and function even after the failure of a drive part of the RAID1 array. It only contributes slightly to throughput performance, as opposed to RAID0, and to a lesser extent, RAID5.

I can't tell if my own practice is used by others for any kind of server, but it works fine for me. I'd rather have all my HDDs and SSDs configured in AHCI storage mode.

But I still capture selective redundancy, save space across multiple drives, and obtain some slight increase in throughput.

Of course, I paid maybe $20 for the privilege in a license to Stablebit DrivePool. I can selectively duplicate files and folders for redundancy; multiple redundant copies of those particular files may improve performance slightly. And if I ever need to recover anything, I can actually do it from any of the pooled drives a disk at a time. DrivePool will tell me when a drive is "in trouble" or failed, and continue functioning until I replace that drive. It will then rebuild the pool from redundant copies. If some files aren't duplicated and are lost on a failed drive, it would've been my choice from the git-go that I wouldn't care that much if they were lost.

But it is just as likely I might recover those unduplicated files from a drive that is "going" bad before all is lost.

And frankly, many of those less-volatile unduplicated files are still backed up on a periodic basis, even if the "period" is a longer one.

 

[grimpr]

RAID 1 provides redundancy, but it isn't "backup." It means that your server would continue to operate and function even after the failure of a drive part of the RAID1 array. It only contributes slightly to throughput performance, as opposed to RAID0, and to a lesser extent, RAID5.

I can't tell if my own practice is used by others for any kind of server, but it works fine for me. I'd rather have all my HDDs and SSDs configured in AHCI storage mode.

But I still capture selective redundancy, save space across multiple drives, and obtain some slight increase in throughput.

Of course, I paid maybe $20 for the privilege in a license to Stablebit DrivePool. I can selectively duplicate files and folders for redundancy; multiple redundant copies of those particular files may improve performance slightly. And if I ever need to recover anything, I can actually do it from any of the pooled drives a disk at a time. DrivePool will tell me when a drive is "in trouble" or failed, and continue functioning until I replace that drive. It will then rebuild the pool from redundant copies. If some files aren't duplicated and are lost on a failed drive, it would've been my choice from the git-go that I wouldn't care that much if they were lost.

But it is just as likely I might recover those unduplicated files from a drive that is "going" bad before all is lost.

And frankly, many of those less-volatile unduplicated files are still backed up on a periodic basis, even if the "period" is a longer one.

Hi, reading that you're a user of Stablebit Drivepool AND Primocache, do you use both on the same server? Are they compatible with each other?

 

[BonzaiDuck]

Hi, reading that you're a user of Stablebit Drivepool AND Primocache, do you use both on the same server? Are they compatible with each other?

Once again -- since I answered your PM -- I was PLANNING to do that. I had a discussion with the Stablebit tech-rep, who didn't see a problem with it, but with this caveat.

Stablebit offers up a "virtual drive" from the pool, but you couldn't cache it with Primo directly, which caches data at the block level. Instead, this is the very simple way it would be done.

You would create a single caching task for all the disk drives in the pool. That is, this single caching task would include all those disks. You should be able to cache at Primo's "L2" level with an SSD allocated for that purpose, then cache all of it to RAM. Or -- you would simply create the task to cache all drives directly to RAM.

For a server OS, you'd use the server version of Primo-Cache. I think that Stablebit offers a single version for various Windows versions, so you wouldn't need a "special license" for that.

BY THE WAY. Another poster touted the ZFS system. Since I can't keep up with everything in my addled late life, I decided to look into it.

There may be advantages to building ZFS pools -- for instance, accommodating multiple file systems. Further, ZFS does its own caching. However, it is recommended to have 1GB of RAM in the system for every ZFS Terabyte.

For me, I like working with what I'm used to if it serves my needs. Once you have your first Stablebit license, the second one is only $15, and Stablebit was my choice when I built my WHS-2011 system. I don't need to work with multiple file systems. And I like PrimoCache's ability to customize the caching tasks, while I don't think that ZFS offers the "L2" SSD-caching feature (why would it?!)

So I'll keep an open mind about this. But if I switch to something new, I want to enter the pool (a pun!) from the steps at the shallow end. Slowly and deliberately -- with a system I choose specifically for my test investigation.

 

[XavierMace]

For me, I like working with what I'm used to if it serves my needs. Once you have your first Stablebit license, the second one is only $15, and Stablebit was my choice when I built my WHS-2011 system. I don't need to work with multiple file systems. And I like PrimoCache's ability to customize the caching tasks, while I don't think that ZFS offers the "L2" SSD-caching feature (why would it?!)

When running ZFS it's typically recommended to run SSD's for your ZIL and L2ARC. The 1Gb per TB recommendation is just a general best practice but can often get away will less depending on your workload. The big advantage of having more is that all your writes hit RAM and the ZIL before going to your slow spinners. So your endpoints only see the time it takes to write to RAM and ZIL which means you get awesome numbers like these:

Plus the other enterprise niceties like dedup, encryption, and snapshots.