Failed RAID Card = losing all data?

[Tomson]

I see this come up from time to time.

"if your raid card dies, you have to get the exact same model otherwise you wont be able to read the data on the hard drives in the array"

The same statement is made for onboard Raid. I know that this isn't always the case, some cards are backwards compatible, but has anyone ever had an experiance of this?

For instance having a 6 year old array + card, and not being able to retrieve the data once the card dies?

Thanks in advance, you can see how this sort of thing gives people sleepless nights though

 

[Dew]

I see this come up from time to time.

"if your raid card dies, you have to get the exact same model otherwise you wont be able to read the data on the hard drives in the array"

The same statement is made for onboard Raid. I know that this isn't always the case, some cards are backwards compatible, but has anyone ever had an experiance of this?

For instance having a 6 year old array + card, and not being able to retrieve the data once the card dies?

Thanks in advance, you can see how this sort of thing gives people sleepless nights though

Personally, I'm not worried. I have three of the same RAID card, so even if one goes tits-up, I have 2 more I can use to recover data.

 

[kleox64]

For onboard raid that is the case, even a BIOS revision will cause issues let alone a new revision in hardware.

For H/W based cards its a non-issue assuming its the same manufacturer and your running a RAID level which is supported on both cards. Currently I'm running a RAID 5 array on an adaptec 5805 (5 series) so a 7 series card (for example which doesnt exist yet!) should recognise the array.

 

[Drudenhaus]

I've seen 3ware card's fail and be replaced with the same model. I have yet to see any of those arrays be trashed from a card failure. It's also worth nothing that it's always best to have the same firmware revision on the replacement card as was on the original card.

 

[AreEss]

It's true and untrue. The following applies to any RAID controller you've ever seen and then some.

If the card is complete freaking garbage, like 3Ware and Adaptec, you absolutely must have the same model and firmware version if you want any reasonable level of safety. This, of course, still doesn't guarantee that your data will be recoverable. 3Ware used to like to try and firmware lock or Sector 0 lock drives to the card, making data completely unrecoverable.

If the card is marginal garbage or designed by idiots, like too many to even attempt listing, the card doesn't matter because your RAID configuration's already gone! Yes, Virginia, some people are so incredibly stupid that they only write stripe to disk, and store absolutely NO configuration data on the disk. Meaning you get a pretty stripe-set of random data, how useful!

If the card is half decent, you'll just need to get a same-model or one that can read configuration data from the disks. (NOT the same as backwards compatible - there is no such thing.) Said card can then, in theory, update the array configuration without data loss. The problem here being that if the size of the configuration data reserved space changes upward? Nope, not happening, unless you want to lose your data.

So in the end, yes, the answer is "invest in a good tape drive or library and take regular backups."

 

[kleox64]

It's true and untrue. The following applies to any RAID controller you've ever seen and then some.

If the card is complete freaking garbage, like 3Ware and Adaptec, you absolutely must have the same model and firmware version if you want any reasonable level of safety. This, of course, still doesn't guarantee that your data will be recoverable. 3Ware used to like to try and firmware lock or Sector 0 lock drives to the card, making data completely unrecoverable.

If the card is marginal garbage or designed by idiots, like too many to even attempt listing, the card doesn't matter because your RAID configuration's already gone! Yes, Virginia, some people are so incredibly stupid that they only write stripe to disk, and store absolutely NO configuration data on the disk. Meaning you get a pretty stripe-set of random data, how useful!

If the card is half decent, you'll just need to get a same-model or one that can read configuration data from the disks. (NOT the same as backwards compatible - there is no such thing.) Said card can then, in theory, update the array configuration without data loss. The problem here being that if the size of the configuration data reserved space changes upward? Nope, not happening, unless you want to lose your data.

So in the end, yes, the answer is "invest in a good tape drive or library and take regular backups."

The adaptec series 5 is the best you will find at this point in time. They are the fastest and most flexible RAID controllers they have ever made, they have solid driver support with multiple OS's and with 24-port versions you really cant go wrong.

 

[Drudenhaus]

If the card is complete freaking garbage, like 3Ware and Adaptec, you absolutely must have the same model and firmware version if you want any reasonable level of safety. This, of course, still doesn't guarantee that your data will be recoverable. 3Ware used to like to try and firmware lock or Sector 0 lock drives to the card, making data completely unrecoverable.

That leaves...LSI and Areca? Why the hate?

 

[Dan_D]

The adaptec series 5 is the best you will find at this point in time. They are the fastest and most flexible RAID controllers they have ever made, they have solid driver support with multiple OS's and with 24-port versions you really cant go wrong.

I too have heard great things about the Adaptec 5 series.

That leaves...LSI and Areca? Why the hate?

I've never really liked 3Ware but Adaptec went from being decent to being almost complete junk over the last few years. Supposedly the 5 series is better, but Areca and LSI are damned good cards and there is little else to bother with out there.

 

[Drudenhaus]

I've never really liked 3Ware but Adaptec went from being decent to being almost complete junk over the last few years.

I was hoping for something a bit more specific than that. I've dealt with 3ware cards at work and home for 3 years now and I can't say I have much on the bad side to say about them (lack of SSD support).

 

[AreEss]

The adaptec series 5 is the best you will find at this point in time. They are the fastest and most flexible RAID controllers they have ever made, they have solid driver support with multiple OS's and with 24-port versions you really cant go wrong.

My 20 years of experience with Adaptec, inside knowledge of their engineering and development is worth far more than your claims that a 24 port card, which is engineered without consideration for electrical requirements is "the best I will find." Period.

That leaves...LSI and Areca? Why the hate?

20 years of Adaptrash. They have always been absolute garbage, the McDonald's of SCSI with worse quality. Do you remember the 1542? No? I do. I remember it's tendency to abruptly drop devices from the bus, it's refusal to handle an in-spec chain with 5 devices or more. I remember fighting 2940UW's for hours upon hours, because Adaptec had decided not to follow SCSI standards, and then attempted to fix hardware with badly written BIOS.
3Ware, we can start with their poor PCI signalling, followed by their atrocious ATA signalling, followed by no less than ten cards which overwrote Sector 0 on the attached drives, so that a vanilla RAID1 disk could not even be detected properly on another system, despite the data on disk being exactly the same as presented.

In other words, in case folks missed it, I've been in the RAID game a very, very long time. I have been privy to a lot of documentation and statistically significant samplings of controllers. It's not my hobby to know these things; it's my job to know the cause and effect of a retryable error, and just what that Sync Rate Reduction will do to my systems.

 

[Dan_D]

I was hoping for something a bit more specific than that. I've dealt with 3ware cards at work and home for 3 years now and I can't say I have much on the bad side to say about them (lack of SSD support).

My experience with RAID controllers isn't as broad as AreEss's is. I will say that I've never used a 3Ware card myself but all the data I've seen on them shows them to be really strong in a couple of areas and really weak in others. They don't have very balanced performance. When I researched cards for purchase I came to the conclusion that LSI and Areca were the way to go. I ended up buying an LSI for my own system.

 

[Dan_D]

My 20 years of experience with Adaptec, inside knowledge of their engineering and development is worth far more than your claims that a 24 port card, which is engineered without consideration for electrical requirements is "the best I will find." Period.



20 years of Adaptrash. They have always been absolute garbage, the McDonald's of SCSI with worse quality. Do you remember the 1542? No? I do. I remember it's tendency to abruptly drop devices from the bus, it's refusal to handle an in-spec chain with 5 devices or more. I remember fighting 2940UW's for hours upon hours, because Adaptec had decided not to follow SCSI standards, and then attempted to fix hardware with badly written BIOS.
3Ware, we can start with their poor PCI signalling, followed by their atrocious ATA signalling, followed by no less than ten cards which overwrote Sector 0 on the attached drives, so that a vanilla RAID1 disk could not even be detected properly on another system, despite the data on disk being exactly the same as presented.

In other words, in case folks missed it, I've been in the RAID game a very, very long time. I have been privy to a lot of documentation and statistically significant samplings of controllers. It's not my hobby to know these things; it's my job to know the cause and effect of a retryable error, and just what that Sync Rate Reduction will do to my systems.

I remember the 1542. I had issues with them similar to what you described. However the 2940UW was pretty good in my experience. The 2940U2W on the other hand was another matter. The 2100DS was junk as well although it was neat looking.

I wish I had the access to the technical data on some of these cards and boards. It would make for a fascinating read.

 

[AreEss]

I remember the 1542. I had issues with them similar to what you described. However the 2940UW was pretty good in my experience. The 2940U2W on the other hand was another matter. The 2100DS was junk as well although it was neat looking.

Only if you got a very late one. I was dealing with them from release. It wasn't till they retooled the whole card until they stopped having a habit of dropping sync rates silently. And even then, the vast majority still do, because they quickly went back to cutting corners.

I wish I had the access to the technical data on some of these cards and boards. It would make for a fascinating read.

If by "fascinating" you mean "utterly horrifying," then yes. If people actually read and understood the documentation on the nForce Ethernet, or many Broadcom chips, or anything from Realtek they'd rather abruptly stop claiming it was "good stuff," to put it mildly.

 

[Dan_D]

Only if you got a very late one. I was dealing with them from release. It wasn't till they retooled the whole card until they stopped having a habit of dropping sync rates silently. And even then, the vast majority still do, because they quickly went back to cutting corners.



If by "fascinating" you mean "utterly horrifying," then yes. If people actually read and understood the documentation on the nForce Ethernet, or many Broadcom chips, or anything from Realtek they'd rather abruptly stop claiming it was "good stuff," to put it mildly.

I know that nForce networking controllers, Broadcom chips and Realtek stuff is pretty bad. However integrated into motherboards its' cheap and good enough for the home user. However anytime I see Broadcom ethernet integrated into servers I cringe.

I wonder if the utterly horrifying stuff you refer to is any worse than reading the AMD processor erratum?

 

[AreEss]

I know that nForce networking controllers, Broadcom chips and Realtek stuff is pretty bad. However integrated into motherboards its' cheap and good enough for the home user. However anytime I see Broadcom ethernet integrated into servers I cringe.

No, no it's not. Wanna know why your system locked up? Because of silicon bugs. Don't think it can happen to you? Guess again. Pretty much the only Ethernet that is not going to bite you in the face these days is NatSemi DP-series or Intel, and good luck getting your hands on NatSemi - I can't get those outside of Suns.

I wonder if the utterly horrifying stuff you refer to is any worse than reading the AMD processor erratum?

Worse than AMD errata. Worse than Intel errata. Worse than Intel Itanic errata.

 

[Dan_D]

No, no it's not. Wanna know why your system locked up? Because of silicon bugs. Don't think it can happen to you? Guess again. Pretty much the only Ethernet that is not going to bite you in the face these days is NatSemi DP-series or Intel, and good luck getting your hands on NatSemi - I can't get those outside of Suns.



Worse than AMD errata. Worse than Intel errata. Worse than Intel Itanic errata.

Ouch.

Well I know Broadcom networking solutions suck ass through personal experience. I even told some higher ups at HP this one time when I used to work for a large customer of theirs.

Intel's network cards are actually my favorite ones.

 

[JeanLuc]

If you find the same card, you can probably get your data back.

 

[Old Hippie]

If the card is complete freaking garbage, like 3Ware and Adaptec,

OK, than how about pointing me in the correct direction for a 2 or 4 port hardware RAID card.

There's gotta be something out there that meets your approval!

 

[Drudenhaus]

OK, than how about pointing me in the correct direction for a 2 or 4 port hardware RAID card.

There's gotta be something out there that meets your approval!

I still like and recommend 3ware and Areca to people. IMO, everything you look into is going to have issues and it's just a matter of how hard you look. Core2 processor errata is pretty ugly, but we all use them!

*puts up Large Shield +1/+4 vs Missles to protect from AreEss' hail of fire*

 

[Dan_D]

I still like and recommend 3ware and Areca to people. IMO, everything you look into is going to have issues and it's just a matter of how hard you look. Core2 processor errata is pretty ugly, but we all use them!

*puts up Large Shield +1/+5 vs Missles to protect from AreEss' hail of fire*

AMD errata is even more horrifying than Intels'. I don't recommend 3Ware but I do recommend Areca. My favorite cards tend to come from LSI however. His comments on Broadcom and Realtek hardware is pretty spot on. As I said good enough for the home user but they are pretty piss poor for server and workstation usage.

 

[Lazn_Work]

my experience is older.. but at work with HP Smart Arrays I have moved drives to a totally different server and had the RAID card come up with good data. (not actually intentionally in that particular case)

The old server was a ML350 G1 with the onboard Smart Array running RAID 5.. The new server was a ML 350 G3 with a Smart Array 6400. The drives were added to be a 2nd array, and they came up with the old servers boot drive etc. visible on it.

However with consumer level hardware this is less likely to happen.

 

[kleox64]

Adaptec 5 series isnt consumer level, they have also found there way into SUN box's this time round.

 

[AreEss]

my experience is older.. but at work with HP Smart Arrays I have moved drives to a totally different server and had the RAID card come up with good data. (not actually intentionally in that particular case)

The old server was a ML350 G1 with the onboard Smart Array running RAID 5.. The new server was a ML 350 G3 with a Smart Array 6400. The drives were added to be a 2nd array, and they came up with the old servers boot drive etc. visible on it.

HP CISS family is all CTD/CBTD. That's Configuration-To-Disk or Backed-To-Disk. That means that even if you blow up the card with a pound of TNT, the configuration data is still there, in it's entirety, and a card that can read the configuration data will bring your array online as if nothing ever happened.

Adaptec 5 series isnt consumer level, they have also found there way into SUN box's this time round.

Just what I needed, another reason to point, laugh and refuse to buy Sun ever again.