RAID 5 Issues

[LIGISTX]

Hey guys, I am not new to RAID, although I am definitely a noob when it comes to it. IE: I know all about how it works and the different variants, but I don't use it much and just have a solid general understanding but nothing in depth like an IT person would who deals with it often.

 

Issue is, I just acquired some Seagate 3 TB drives *not RAID specific, just consumer drives* 7200.14's, and I decided hell, I will upgrade my old RAID solution on my HTPC that was pretty crap. I had to update my mobo's BIOS since it wasn't happy at all about anything, so nbd. Deleted my old RAID after moving my data to an external, set up new RAID, and away we go. Load into windows 10, Intel RST shows evrything is cool, start the copy, and first issue is I am only getting ~25 mbps. So that kinda sux, but I figure the intel RAID controller on my Z67 board isn't top notch, so thus is life. I let it go overnight, only to wake up to a FAILED RAID array. The lowest drive in Intel RST showed as dropped out. So I decide maybe something weird happened, so I cancel the transfer, delete and redo the RAID just to be safe, and this morning it happened again, lowest drive in Intel RST was dropped out.

 

So I am not sure how I RST works, but I assume this means the same drive failed, I will have to investigate more later when I get home from work to see if the same drives always show up in the same order in I RST based on SATA port on the mobo, if so, the same drive has dropped out twice in 1.5 days.... I have a few RAID setups and have never had this issue, so I am not sure what is causing it. Only thing I can think of is the drive being bad. This mobo had a RAID 5 on it before and it never had a hiccup, and if I do remember right when these drives where not in RAID before in my parents PC, once or twice over a year span one of the drives was no longer showing in Windows - I assume it was this drive since its having issues.

 

Would a possible firmware update help? Or is there just something wrong with the drive? And if so, being a consumer drive would Seagate even warranty it if I say it has RAID issues, probably not. At least I can fall back on it just disappeared in Windows while not in RAID...

[Bittenfleax]

@{EAC} Shoot em UP Test the drive on its own? Do some benchmarks compared to the other ones and see if it is slower/faulty. If not, firmware upgrade.

[LIGISTX]

@{EAC} Shoot em UP Test the drive on its own? Do some benchmarks compared to the other ones and see if it is slower/faulty. If not, firmware upgrade.

I could try that, but I am thinking I will just try and RMA it. It dropped out of normal windows environment on me a few times but I never thought anything of it, but now after thinking about it more that tells me there has to be an issue with the drive. I would try firmware first, I suppose I will see if they have new firmware for the 7200.14's. I have A LOT of them deployed in RAID and not in RAID environments and this is the first issue I have had, so I don't think its firmware since most of them are all on the same or older firmware... But worth a shot.

[Bittenfleax]

I could try that, but I am thinking I will just try and RMA it. It dropped out of normal windows environment on me a few times but I never thought anything of it, but now after thinking about it more that tells me there has to be an issue with the drive. I would try firmware first, I suppose I will see if they have new firmware for the 7200.14's. I have A LOT of them deployed in RAID and not in RAID environments and this is the first issue I have had, so I don't think its firmware since most of them are all on the same or older firmware... But worth a shot.

Yeah fair enough.

[LIGISTX]

I guess the main reason for this thread was to find out if most of the time it is just a bad drive. Or do consumer drives "tend" to fall out of RAID arrays? I assume not.... I am pretty sure that isn't a normal thing to happen and there must be something wrong with the drive to cause it to just quite talking to the RAID controller and cause it to drop. Putting the computer to sleep and waking it fixed it which is interesting.

[Vitalius]

@{EAC} Shoot em UP

For your speed issue, yeah that will be your controller. Remember that your CPU has to calculate parity on-the-fly while doing parity RAID. It's a potential bottleneck too.

This is why I vastly prefer RAID 10. No parity to calculate. Much more speed and equal to or better redundancy. You just get less space (which most people consider the deal breaker). 

I was looking for Linus' video on RAID 5 vs RAID 6, but I can't find it (may have forgotten the title).

TL;DR: With larger HDDs, RAID 5 is bad because an Unrecoverable Read Error will make a drive drop out of the array even if the rest of the drive is fine. 1 bit being unreadable makes the rest useless (in a RAID anyway). 

This may not be your problem, but that one drive you have could be having this issue where it's unable to write/read from a portion of the disk.

Usually, if your RAID array is failing, it isn't just 1 drive with this issue since RAID 5 is redundant for 1 drive failure. It's usually two or more that have it and cause a RAID array to fail. 

It's a sinister issue because it isn't cut and dry. It could work once, you think everything is fine, then it fail again once you've started using it. 

A drive dropping out and the RAID failing after it tried to rebuild or you dumped a ton of data on it is what happened to me at my workplace with our 7 drive RAID 5 array. One drive would say it failed, and the rest would seem fine, but it was always a different drive (this is likely because we had 7 drives in the array which makes that more likely to happen). 

Anyway, I'd either RMA or at least check the SMART data on the failing drive(s). It usually lists UREs (Unrecoverable Read Errors) that occur or something like it. Super high seek time is the issue basically. 

To give more regarding that, when your drive has a URE occur, it keeps seeking that bit of data, unable to get it. Your RAID controller is waiting on the drive, but the drive is still trying to find what it needs. Eventually, your RAID controller gives up on waiting (like 45 seconds usually?) and just goes "your array failed". 

Western Digital Reds and other large/enterprise storage oriented drives have shorter "we give up" times set in them for UREs so that they don't cause this. i.e. a single file is corrupt, but the rest are fine and accessible, as compared to the whole drive being considered "failed". Regular consumer drives have longer seek times for UREs before giving up because it's more important for regular consumers who don't use RAID to get that bit of data back as they are far less likely to have backups. 

You could just use RAID 6, because the extra redundancy will basically prevent this from happening (two drives have to have a URE or fail in RAID 5 for it to happen, and that's somewhat likely, but three drives have to do this for the RAID 6 to fail, and that's much less likely). 

It's just a natural issue with RAID 5. Only 1 drive of redundancy and you are using 3TB drives (bigger the drive = more likely to have a URE). There's no way around it other than using different HDDs or a different kind of RAID.

[EmeraldFlame]

I guess the main reason for this thread was to find out if most of the time it is just a bad drive. Or do consumer drives "tend" to fall out of RAID arrays? 

 

Consumer drives to tend to fall out of RAID arrays. Basically if a read/write request takes longer than normal and the drive doesn't report back the raid controller assumes the drive isn't responding so it drops out. 

RAID drives get around this with a function thats built into the firmware that's called TLER (I think seagate has a different name for it). The shortest explanation is it basically allows the drive to say 'hey I'm here, I'm just getting some read errors and it's taking me a little longer than normal, don't drop me'. It's obviously more complicated than that, but you get the idea. 

[LIGISTX]

@{EAC} Shoot em UP

For your speed issue, yeah that will be your controller. Remember that your CPU has to calculate parity on-the-fly while doing parity RAID. It's a potential bottleneck too.

This is why I vastly prefer RAID 10. No parity to calculate. Much more speed and equal to or better redundancy. You just get less space (which most people consider the deal breaker). 

I was looking for Linus' video on RAID 5 vs RAID 6, but I can't find it (may have forgotten the title).

TL;DR: With larger HDDs, RAID 5 is bad because an Unrecoverable Read Error will make a drive drop out of the array even if the rest of the drive is fine. 1 bit being unreadable makes the rest useless (in a RAID anyway). 

This may not be your problem, but that one drive you have could be having this issue where it's unable to write/read from a portion of the disk.

Usually, if your RAID array is failing, it isn't just 1 drive with this issue since RAID 5 is redundant for 1 drive failure. It's usually two or more that have it and cause a RAID array to fail. 

It's a sinister issue because it isn't cut and dry. It could work once, you think everything is fine, then it fail again once you've started using it. 

A drive dropping out and the RAID failing after it tried to rebuild or you dumped a ton of data on it is what happened to me at my workplace with our 7 drive RAID 5 array. One drive would say it failed, and the rest would seem fine, but it was always a different drive (this is likely because we had 7 drives in the array which makes that more likely to happen). 

Anyway, I'd either RMA or at least check the SMART data on the failing drive(s). It usually lists UREs (Unrecoverable Read Errors) that occur or something like it. Super high seek time is the issue basically. 

To give more regarding that, when your drive has a URE occur, it keeps seeking that bit of data, unable to get it. Your RAID controller is waiting on the drive, but the drive is still trying to find what it needs. Eventually, your RAID controller gives up on waiting (like 45 seconds usually?) and just goes "your array failed". 

Western Digital Reds and other large/enterprise storage oriented drives have shorter "we give up" times set in them for UREs so that they don't cause this. i.e. a single file is corrupt, but the rest are fine and accessible, as compared to the whole drive being considered "failed". Regular consumer drives have longer seek times for UREs before giving up because it's more important for regular consumers who don't use RAID to get that bit of data back as they are far less likely to have backups. 

You could just use RAID 6, because the extra redundancy will basically prevent this from happening (two drives have to have a URE or fail in RAID 5 for it to happen, and that's somewhat likely, but three drives have to do this for the RAID 6 to fail, and that's much less likely). 

It's just a natural issue with RAID 5. Only 1 drive of redundancy and you are using 3TB drives (bigger the drive = more likely to have a URE). There's no way around it other than using different HDDs or a different kind of RAID.

 

Yes, this all makes sense. Except for the speed of the CPU issue, this is on a 3770k lol. I fully understand the SATA controller isn't exactly top notch here, but the CPU sure shouldn't be the issue, I admit the communication lines of the mobo are not set up for super RAID 5 performance though, so that is fine and I accept it.

 

I may try RAID 10 (if my mobo even supports it, it should) as your right, this would solve all the speed issues, although in HDtune I was getting like 500+ mbps read out of the RAID 5 which is plenty fine, but 25 mbps writes kinda suck, although I feel like 25 is REALLY low, maybe has to do with this drive being bad and its seek times being all over the map causing the entire array to slow down. I suppose I will delete the RAID once more and see if there are any smart errors on any of the drives after loading them up with a bunch of DATA, my main drive in that rig is an SSD, so I will just copy a bunch of data onto all 4 drives at once from my SSD and see what happens.

 

I am thinking RMA, but I will try this first I suppose.

Consumer drives to tend to fall out of RAID arrays. Basically if a read/write request takes longer than normal and the drive doesn't report back the raid controller assumes the drive isn't responding so it drops out. 

RAID drives get around this with a function thats built into the firmware that's called TLER (I think seagate has a different name for it). The shortest explanation is it basically allows the drive to say 'hey I'm here, I'm just getting some read errors and it's taking me a little longer than normal, don't drop me'. It's obviously more complicated than that, but you get the idea. 

Correct, I agree with this, but I feel like it shouldn't happen 100% of the time lol