hardware (HDD) rant

[meton_magis]

I'm so mad right now, I built a computer specificaly to be a server (mostly squid and fileserver.) It had 3 1.5TB Seagate Barracuda drives in a raid 5 array. I'm doing some programming on it (I use it as a dev environment when I'm away from home) and I get a message that 1 of my disks failed. I check it out, and sure enough, 1 of my drives fell out of the array. No big deal, that's why I spent money to make the array, and the disks are still under warranty. I'll just go home, and call up support to RMA the drive, a week later I'll have a replacement that I can drop in, rebuild, I'm good to go. Less than 1 hour later, a second drive dies, and I lose the entire array. . Luckily I had only around 300 GB of data on the array, and I had just reinstalled Fedora 11 on there, so my last backup to an external hard drive was fairly recent. I'm a bit scared though since my exhdd is a seagate ><

That'll be the last drive I buy from seagate though ..... especialy after reading about their giant firmware issues

Seagate Barracuda - Wikipedia, the free encyclopedia

I'm not sure if this was the cause of my failures, but having 2 drives fail THAT quickly when there was barely any load on them (programming, not compiling) ..... ugh.


</rant>

[darkrose0510]

At least you didn't lose much, I once lost a fairly big chunk of some libraries I'd been working on due to a seagate drive dieing, I only buy Western Digital these days.

[Rubberman]

I've generally had very good results, reliability wise, with Seagate drives. However, like most manufactured things, problems often happen in batches, which is why when you get a bunch of drives from the same manufacturing batch and then use them in a RAID device, if one fails it will often happen that one of the others will within a short period of time. This is true of all drive manufacturers, not just Seagate. You were just unlucky and got the short stick, unfortunately.

My recommendation is when building RAID devices, make sure the drivers are not of close sequential serial numbers (from different manufacturing batches), and make sure you are dilligent in making full backups during the burn-in period of the drives (3-6 months) when they are most susceptible to failure.

[coopstah13]

raid is not foolproof, like rubberman said drives usually fail together from the same batch and you probably got some from the same batch, make regular backups

also, with disk space being so cheap it might make sense to add the extra parity disk and go raid 6

[Lensman]

The first HD I ever owned was a 20M Seagate that I bought years ago. For all I know, that drive is still working. Since that time, I have mostly owned Seagate drives and I have been quite pleased with them. Lately, however, I have switched most of my drives over to WD, as I believe they are more robust than the current batches of Seagate drives. That said, I would never trust any software RAID regardless of the drives used.

[questio verum]

If you don't mind me asking, were these OEM or retail packaged drives? If they were OEM, who was the source? (you can just list the initials if you don't want to name them).

I've heard anecdotally that Seagate had fallen off a little on their QA. I hope they straighten things out... I've always had good experiences with their drives in the past.


qv

[a12ctic]

Been using Samsung drives lately. The F3 is a very fast drive, is affordable, and has 500GB layers.

[meton_magis]

If you don't mind me asking, were these OEM or retail packaged drives? If they were OEM, who was the source?

They were OEM drives I bought from tigerdirect.

also, with disk space being so cheap it might make sense to add the extra parity disk and go raid 6

I guess at the time I figured that if 2/3 disks can go bad, why would 3/4 be much better off?

I still have this opinion, I just never though of the above mentioned flaw of buying 3 disks at once that are probably of the same batch.

next time I buy for a raid (I'm going to RMA these, and try to use new ones, just paying better attention to backups.) I'm probably going to go 1 samsung, 1 WD, 1 hitachi (they sound cheap, but I had 3 hitachis that I bought several years ago that are still going strong.)

That said, I would never trust any software RAID regardless of the drives used.

Why? I've never heard of raid failures due to the linux raid drivers not working. My issues were not an issue with the raid software (in fact everything worked perfectly fine until I lost the second drive.) A "hardware" raid solution is no different from a "software" raid solution other than the cpu / bus / ram is on a second PCB that is attached to the computer, who's "software" to run the raid is usualy, if not always, proprietary.

I've never understood what people have against software raid, especialy in a linux community where for whatever reason, we've chosen to use free software.

[coopstah13]

if anything, a software raid is safer, because as long as you have a linux distro that has mdadm and can understand it, you should be able to use it

with a hardware raid, you would need a nearly identical raid card, what if your model is discontinued?

The only real difference is that a hardware raid controller usually has its own processor to handle the calculations rather than relying on the CPU of the machine.

[meton_magis]

if anything, a software raid is safer, because as long as you have a linux distro that has mdadm and can understand it, you should be able to use it

with a hardware raid, you would need a nearly identical raid card, what if your model is discontinued?

The only real difference is that a hardware raid controller usually has its own processor to handle the calculations rather than relying on the CPU of the machine.

That is exactly my thoughts on the issue. Less hardware == less things to fail. Software can be recreated (the driver to understand raid) but a PCB has to be replaced, and unless I want to recreate data, I'm stuck with vendor lockin.

Even in an enterprise type environment, where I'd NEED a dedicated CPU to handle raid, I'm more likely to get a SAN (or is it NAS, I confuse the 2) that connects directly to the machine to handle storage, where the machine IS the raid CPU.

EDIT: I was thinking of a SAN, although a machine dedicated to storage serving through NFS and samba is a solution I use quite a bit.