Swap out LSI card for Adaptec - Will I lose data

[The_Tango]

I am thinking about swapping out a LSI MegaRAID SAS 84016E for a Adaptec RAID 52445 card. I was wondering if I will lose my data (Raid 6)? If I power down , swap cards, configure Adaptec for Raid6 - Will it see the data that the LSI card wrote?

I really have too much data to do any sort of backup (over 12TB).

 

[Hydaral]

I have not had any actual experience, but from everything I have read, even swapping between different models of RAID card from the same manufacturer will not be compatible. So I think you should assume that you will need to recreate your array, unless someone else can confirm that you can switch between controllers without data loss.

This is one of the major drawbacks of hardware RAID.

 

[stevech]

Isn't it the case that common low cost or motherboard RAID1 mirroring yields two identical drives, that in Windows, are both usable standalone (NTFS)?

 

[BriComp]

I would expect that going LSI to Adaptec and you will lose your data ( that is just unplugging LSI and plugging in Adaptec ).

I have found with a Raid0 setup that it is possible to Unplug an LSI card and re-plug a better LSI card without losing data. It s not possible to go from a higher spec to a lower spec card.

By not having a back-up of your data you seem to be running fast and loose with it anyway. Raid6 is NOT a back-up solution.

 

[The_Tango]

I would love to have a backup - But how do you backup 12TB of data without building another box?

Online isn't practical, would take months if not years. Tape - I would be knee deep in tapes.

Raid 6 seemed the closest thing. At least two drives can fail (at the same time) before I'm in trouble.

I'm open to any suggestions that don't cost a fortune.

 

[sammynl]

Array metadata is NOT on the RAID controller, but always on the disks (persistant superblocks). So if you attach raid array member disks to a new identical controller the RAID array will remain intact. The controller will find a foreign array and most likely (depending on firmware) will ask you if you want to import the existing foreign array (yes) or if you want to build a new array (no). Do keep in mind to label your disks (0 1 2 etc..) and connect them in the same sequence to your controller (disk 0 - port 0 disk 1 port 1 etc.)

However this all depends on signatures, both on the array disks and on the controller. If they match the controller will see the array as native and not foreign and simply mount it. If the signatures don't match, the controller will offer you to import the foreign array and mount it based on the metadata (as raid 0 / raid5 or whatever).

Problem is that if you attach the disks to a different type of controller the signature cannot be read by the controller in the first place, hence it cannot be recognised as foreign and therefore cannot be imported.

 

[The_Tango]

Thanks sammynl for a easy to understand explanation.

The good deal I was going to get on the Adaptec RAID 52445 card fell through. So I will NOT be swapping out the cards.

 

[stevech]

So a person with a lot of data in a RAID array should have a spare identical RAID controller as a CYA measure, eh?

Aren't there any standards in work so controllers can be interchanged, esp. if one goes end-of-life in availability?

 

[chiadog]

One can always opt for linux RAID if they haven't invested in a proper RAID card (or toss it like I have). The portability between different kernel version may or may not work, but you can just keep a copy of the exactly bootable version on a thumb stick to recover the array if need be.
But to directly address the question, I would say depends on how much your data is worth to you. Buying an spare card is just insurance just in case you cannot source another one down the line when it cough up the ghost. Having a spare still doesn't mean you can't lose the array, either due to technical issues or ID 10 T errors (my case). I guess the lesson is: always have multiple backups. That's probably more important than having a spare card.

 

[Hydaral]

So a person with a lot of data in a RAID array should have a spare identical RAID controller as a CYA measure, eh?

That, or a recent backup.

 

[sammynl]

So a person with a lot of data in a RAID array should have a spare identical RAID controller as a CYA measure, eh?

RAID is not back-up, thinking it is will sooner or later result into a 100% dataloss. A good back-up is the only assurance you won't loose data, not RAID i'm afraid.

 

[sk1939]

Sometimes it's not practical to backup that much data however. I know that at one place I worked at only backed up critical files, and even that was several TB a week. We had redundant on and offsite arrays for "backup" in addition to the tape. There is no good way to backup a couple hundred TB's of data.

 

[sammynl]

There is no good way to backup a couple hundred TB's of data.

That requires a NAS/SAN with (real-time) mirroring, meaning you have a read-only copy of all you data on a second NAS/SAN. Than snapshot the mirror and back-up the snapshot. In a NAS environment you might consider snapshots as well not to help you in case of hardware failure but to go back to any point in time depending on your snapshot schedule. In general these amounts of data are on SAN's with archivers were static data is moved to different disks/array's based om some policy, that data can be back-upped within a large time-frame and the dynamic data in a small time frame. The archived data can always be accessed by clients because the archiver creates pointer files (8Kb only) on the active array allowing you to recall files at any time. Another workable solution (cheaper) is the use of incremental back-up's daily monday-friday and a full backup once every weekend (large timeframe) using tape-robots. For the homebrewed NAS you can make incremental back-ups with robocopy