0

I have a server with Adaptec HostRaid controller. It has 2x 2TB drives. If I go to create a RAID0 array (striped), it will set the array's capacity to 2TB, when I go to create a RAID1 array (mirrored, and what I want to use), it sets the array's capacity to 1023GB instead of something closer to 2TB. The utility shows both drives as 1.8TB, what do I need to do to create a ~1.8TB RAID1 array?

SuperMicro321
  • 41
  • 2
  • 4
  • Update: I upgraded the BIOS to the latest version (From 2008, previous was from 2006) and I'm still having this issue. The motherboard is a SuperMicro X7DBR-i+, BIOS is the latest 7DBR R 2.1a. Adaptec RAID Configuration Utility V5.0-0 B229. The motherboard also offers Intel Matrix FRAID, but that fake raid doesn't work with CentOS. If I go to create 1 RAID1 (striped) array, which is not what I want, it says the array capacity will be 2TB... – SuperMicro321 Jan 20 '11 at 10:36
  • Can you clarify the above? You say "if I go to create 1 RAID1 (striped) array, which is not what I want, it says the array capacity will be 2TB", but that is in fact the right size and what you want, unless I misunderstand. Do you mean that if you go to create 1 RAID*0* array, etc. etc.? Also, and forgive the question, but can you read the model numbers of the drives, and paste them into your answer? You give impressive detail on everything except that. – MadHatter Jan 20 '11 at 10:41
  • My mistake, RAID0 striped. Drives are both Hitachi 2TB SATA HDS722020ALA330 – SuperMicro321 Jan 20 '11 at 10:44
  • I went ahead and began to build the RAID1 array, it now says that the array will be 1.8TB. Weird, but I'm crossing my fingers that it's right! – SuperMicro321 Jan 20 '11 at 12:31

4 Answers4

1

This is the typical distinction in how disk manufacturers and logic systems measure disk space, disk manufacturers will assume that 1Tb is 1000 Gb where most of the other systems it'll assume that 1Tb is 1024Gb.

Apart from this consideration it looks like your RAID card isn't performing its job, it could be highly possible that it doesn't see all the hard drive sectors, since the introduction of drives with 4k per sector (normally it was 1k per sector) most hardware devices need a new firmware update to support this, check with Adaptec to see if you can update your firmware and check if that helps, you should end up with a RAID1 with a capacity of 2Tb (or 1.8Tb, same same).

lynxman
  • 9,397
  • 3
  • 25
  • 28
1

Creating a RAID 0 Array of 2x2TB drive should give you a resulting array of around 4TB, which it is not. I would suspect that your RAID controller has issues in seeing drives greater than 1TB in size. I would look for a firmware update for your controller to allow it to make use of the full capacity of all of your drives.

Sam Cogan
  • 38,736
  • 6
  • 78
  • 114
1

OK, no worries, and thanks for the drive models.

I have to agree with the second part of lynxman's answer, that limit of 1023GB sounds very suspiciously like a programming issue with the RAID controller, and if bringing it as up-to-date as possible doesn't give you all 2TB, then I fear using software RAID will be your best bet.

CentOS 5 supports an install onto software RAID, so you shouldn't have any problems with that.

For what it's worth, I personally prefer software RAID to shonky hardware RAID anyway. An important element of RAID on a server is the ability to know when a drive has failed - otherwise you generally only find out when the second drive has failed, and that's not such great protection. Whilst separate RAID cards often have good Linux support, I've found that many short-lived on-the-motherboard hardware controllers have very poor Linux support, and offer you no way to check the state of the underlying drives from the running OS. mdadm will definitely give you fine-grained control over your RAID, plus the ability to use smartctl to schedule in-service tests on the drives, and hopefully to predict their eventual failures, and gracefully deal with that.

MadHatter
  • 79,770
  • 20
  • 184
  • 232
  • Your arguments may be valid, but todays' RAID controllers offer SMTP and SNMP notifications in case of a failure. Or servers with ILO send alarms via SNMP/IPMI etc so you won't miss a thing. – weeheavy Jan 20 '11 at 12:41
  • I work at the DC that this server will be colo'd at, so I'll be able to see a red light on a bad drive too :P – SuperMicro321 Jan 20 '11 at 12:59
  • I'm not knocking either of the above, but to weeheavy I'd add that SNMP traps are only any help if you have an SNMP trap infastructure set up, and I'm not aware of anything lightweight and free that does that (keen to learn, though); as for SMTP, does the RAID card have its own ethernet port, or does that require the cooperation of the OS? If the former, then it's definitely not a shonky hardware RAID, and as I said, my hardware RAID issues are only with the shonky end of the market. – MadHatter Jan 20 '11 at 13:32
0

The numbers for the RAID1 setup are correct: 1 Terabyte is only 1,000 Gigabytes, etc...

The reason you see a neat 2048Gigabytes for the RAID0 array is because your RAID card cannot cope with anything (hard disk or RAID array) bigger than 2048Gigabytes. So don't setup the RAID0 with its current BIOS/firmware: it will crash and burn at some point, a bit like a USB stick with fake capacity.

DutchUncle
  • 1,265
  • 8
  • 16