Ubuntu 12.04 won't see raid 0

[MAFoElffen]

That's exactly what I meant. mdadm is very easy to make it work. Where did it fail?

A dedicated raid card might seem like the easy way out, but apart from additional cost you are also tying yourself up to a particular hardware card. For example, if it dies, you need to use the same model to get your array working and even then it might fail.

With SW raid you can read it on any linux machine, not only ubuntu.

But at the end of the day, the choice is all yours (or your boss's).

+1 with darko

Some hard-RAID cards are a pain. Some make things transparent and easy. I have an IBM XSeries server... If I have a problem with Hard-RAID, I have to boot off the CD to run the config utilities for it. The on-controller BIOS util is just a status, see what's going on affair. PITA.

In the server world, I think LVM is handy if I keep it on hardware RAID, have a DOM0 host to DOMU virtual hosts that I am leasing out. (And they change in size requirements or go away.) The only handy part about it is that you can extend and shrink a logical volume, adjust the size of the file system, add disks to it or take away, delete volumes, etc... very flexible, changes quickly and on the fly.... But also a bit touchy, finicky and not for the light hearted. They don't do recovery very well. Yes, you can take a snapshot and the backup LV will track every change from that snapshot... Myself- I use LVM for tests(Within a physical volume on RAID), but I don't trust it's hardiness for production. The physical volume or logical group goes down and you're screwed. MDADM is better for that and IMO more flexible.

So the big difference in admin speed is it takes under a minute to double the size of a Logical volume, where it seems to take forever to grow a RAID array or for it to rebuild. But that is just it, it can rebuild dependably...

You said you had onboard. Was your onboard- HP Smart Array P420/1GB FBWC Controller (RAID 0/1/1+0/5/5+0)? If so, that is hard RAID... That was an option card for that server. The other 3 where Dynamic Smart Array controllers with RAID... but I'm not familiar with those. I do know that they only did RAID)0 and RAID1- unless you bought a 512MB cache upgrade to do RAID5 on them. Other than that, I don't think that server board bios itself had a 'fake-raid" in it's bios... But I don't have that particular "Generation"... Out of the 10 test servers here at home, one of mine is a HP Proliant DL380 G3... So it's a bit older.

Does your Controller card BIOS have in-BIOS configuration? I'm wondering if it has the drivers to see the logical drive in your card... as I see hardware drivers and software utilities for Suse and RedHat... And I don't know how lspci ID's that card... to check if there is a kernel driver for it...

On hard-raid, pass-through drives in setup are called "JBOD", which stands for "Just a Bunch Of Disks".

I do RAID. Mostly RAID 5 or 5EE.

[MAFoElffen]

I guess I should translate my last post and explain... Ubuntu is pretty good at finding common hardware RAID devices without a user having to load additional external drivers during the install. (Usually) Most of the time this goes along as transparent during the install. A lot of "common" controllers are opensource and so available... Sometimes not and then you have to pop out to a shell to install a driver for the system to "see" it.

Easy is not true for others Distro's. Example- Loading Solaris or Oracle Linux (Both RHEL) on my same boxes and I always have to load drivers for the system to see those Hardware RAID drives... unless I pass them "through" as not RAID (JBOD) for the system to see without those specific drivers. Just something to think on. You still have to set them up as that. In RAID, for RAID type, instead of RAID0 or RAID1 (etc), usually JBOD is there as a RAID type.

Example, today setting up an IBM XSeries server. In Solaris, I start and it says it has no drives. I load their Live Image, it says the ServeRAID driver is missing, then asks for a repo or drive path for the driver for it to install. What I used in that was that it gave me the driver package "name" of what it needed for that device. That way I could make sure I have that driver package for my install. Of course in a text install, you have to pop to a shell to do that manually. I usually have drivers for a specific box copied on to a USB thumb-drive and keep that for each server. Oracle Linux Server r6, same way. Have to load the specific RAID Card driver for the system to see the RAID logical drive. Same box, Ubuntu see's that logical drive fine in the Partitioner... but it did "find" a driver for it when it did the hardware probes...

Like I did, a way to work things out on a system is too use that Distro's LiveCD image (Your case Ubuntu) to use for diagnostics and to be able to work out the details of what you need and how to make it work. Get the driver working in the LiveCD image. Then have it locally available to use for the install.

Like darko said... Unless you 'really" need that for production / real world commercial support, mdadm is a lot easier to setup.

[offerlam]

Hi Darko and MaFoElffen,

Thanks for you input MaFoElffen!!

but

My boss is persistant in wanting to use a controler despite me pointing at the screen on your inputs and singing the song of easy recovery if we have a failure and all that jazz ...

So taking what MaFoElffen says i would need to find a controller that supports raid 5 and which Ubuntu server 12.04.02 has drivers for. Can you recommend any? Been looking at the compatibility list and search the forums but couldn't find any list for Raid controllers?

[darkod]

Something like this?
http://www.ubuntu.com/certification/...category/RAID/

But that list is very short, I'm sure it works on many more components. Since the boss insists, you can consider one of the LSI models.

[MAFoElffen]

You know there is utilities and tutorials on converting rhel packages to .deb (debian) right? So you can convert the drivers for what you have to a debian package or compile them from source. But if your card doesn't have the option module for RAID5... Yes would recommend a better card.

I do mostly Linux and UNIX server support.

My recommends- Adaptec, then LSI...

I love LSI cards! I've always felt that they are the industry standard for RAID cards. They always seem to be the one that set the standards in new technology that everyone else then copies or emulates. Sidenote on that list- 3ware, MegaRaid & LSI are the same entity now.

But for HP Servers, I lean towards the Adaptec Cards. Reason why? "Some" HP servers have something different about their BIOS keyboard interrupts. In the past, I've had some problems (with some boxes) putting LSI cards in HP servers. They come up fine, but then when you go to hit their BIOS hotkey to configure them, the Main system's BIOS intercepts it (for their use) causing some kind of conflict and you can't get into the RAID card's BIOS. You can still get in after it boots via software configuration management... But if you do have a problem like that, it's hard to "disaster manage" that server if it doesn't boot to a OS prompt.

But like I said, that is only with some HP Servers with some LSI cards. There was a LSI firmware microcode update that was supposed to address that. But after burning those update images to card BIOS, I still have some LSI cards that still don't work with some of my HP Servers. So if you try an LSI card on it, make sure you can return it if it ends up having that conflict problem. That conflict doesn't have anything to do with how it functions, just in being able to manage it through the hotkeys. Those same cards of mine are working just fine in other servers. I think the LSI cards have more features, but they have to work with what you have right? LSI cards come with drivers for rhel Linux and source that you can compile for debian... Although they say that, I haven't had one not seen by debian or Ubuntu, so they seem to be covered by the Ubuntu repo's drivers. But... read on...

Adaptec has a reputation of being more open to "opensource" with their driver code to Linux and UNIX. And there are so many Adaptec cards out there working on everything... Their Series 6 and Series 7 cards are directly supported with debian and "ubuntu" drivers and software packages from them through Ubuntu 12.04LTS (so far):
http://www.adaptec.com/en-us/products/
For a production box I would go with their Series 7 cards, for end-of-life factors... It is what is their current production line. Their older 2 through 5 series cards came with Ubuntu drivers, but they stopped direct "flavored" driver download support from them for those cards at Ubuntu 11.04. I still think those older cards are supported directly by drivers in the Ubuntu Repo's, by AAC.

Whatever you get, 64bit PCIE, battery backed BIOS... able to configure with both the onboard BIOS and the software utils... Alerts through BIOS or software monitored. Able to do RAID5, RAID5EE and maybe even RAID6. As with other system alerts, I have my RAID status alerts (both mdadm and hardware RAID), SMART status and storage capacity alerts emailed to "me" from my systems.

[offerlam]

You know there is utilities and tutorials on converting rhel packages to .deb (debian) right? So you can convert the drivers for what you have to a debian package or compile them from source. But if your card doesn't have the option module for RAID5... Yes would recommend a better card.

I do mostly Linux and UNIX server support.

My recommends- Adaptec, then LSI...

I love LSI cards! I've always felt that they are the industry standard for RAID cards. They always seem to be the one that set the standards in new technology that everyone else then copies or emulates. Sidenote on that list- 3ware, MegaRaid & LSI are the same entity now.

But for HP Servers, I lean towards the Adaptec Cards. Reason why? "Some" HP servers have something different about their BIOS keyboard interrupts. In the past, I've had some problems (with some boxes) putting LSI cards in HP servers. They come up fine, but then when you go to hit their BIOS hotkey to configure them, the Main system's BIOS intercepts it (for their use) causing some kind of conflict and you can't get into the RAID card's BIOS. You can still get in after it boots via software configuration management... But if you do have a problem like that, it's hard to "disaster manage" that server if it doesn't boot to a OS prompt.

But like I said, that is only with some HP Servers with some LSI cards. There was a LSI firmware microcode update that was supposed to address that. But after burning those update images to card BIOS, I still have some LSI cards that still don't work with some of my HP Servers. So if you try an LSI card on it, make sure you can return it if it ends up having that conflict problem. That conflict doesn't have anything to do with how it functions, just in being able to manage it through the hotkeys. Those same cards of mine are working just fine in other servers. I think the LSI cards have more features, but they have to work with what you have right? LSI cards come with drivers for rhel Linux and source that you can compile for debian... Although they say that, I haven't had one not seen by debian or Ubuntu, so they seem to be covered by the Ubuntu repo's drivers. But... read on...

Adaptec has a reputation of being more open to "opensource" with their driver code to Linux and UNIX. And there are so many Adaptec cards out there working on everything... Their Series 6 and Series 7 cards are directly supported with debian and "ubuntu" drivers and software packages from them through Ubuntu 12.04LTS (so far):
http://www.adaptec.com/en-us/products/
For a production box I would go with their Series 7 cards, for end-of-life factors... It is what is their current production line. Their older 2 through 5 series cards came with Ubuntu drivers, but they stopped direct "flavored" driver download support from them for those cards at Ubuntu 11.04. I still think those older cards are supported directly by drivers in the Ubuntu Repo's, by AAC.

Whatever you get, 64bit PCIE, battery backed BIOS... able to configure with both the onboard BIOS and the software utils... Alerts through BIOS or software monitored. Able to do RAID5, RAID5EE and maybe even RAID6. As with other system alerts, I have my RAID status alerts (both mdadm and hardware RAID), SMART status and storage capacity alerts emailed to "me" from my systems.

what would you prefer to use in a server production environment? Hardware Raid Adaptec or Linux (Ubuntu server) Software raid?

[offerlam]

Guys I'm in all secreacy trying to get software raid to work... don't tell anyone my boss won't know the difference anyways.. i want this to work..

what i have done so fare is that i have bootet into the ubuntu server install, NOT live cd, and made the following during the patitioning fase:

made one partition marked as boot and made bootable 2Gb
made one partition on all the drives 1 through 4 marked for swap 2Gb
Made one raid5 4 disk 0 spare marked as EXT4 with root partition

When i get to the Grup boot loader is stalls and says

An installation step failed. you can try to run the failing item again from the menu or skip it and choose something else. The failing step is: Install the grup boot loader on hard disk

What do i do? im stuck?

[darkod]

If I remember it corectly (reading many threads here), you are using 3TB disks with gpt table which means you also needs a small 1MiB UNFORMATTED partition with the bios_grub flag. Grub2 can't install on gpt disk without it. I guess that's the reason it's failing.

See one of my previous posts for the suggested disk layout/partitions. I think it was in this thread.

[MAFoElffen]

what would you prefer to use in a server production environment? Hardware Raid Adaptec or Linux (Ubuntu server) Software raid?

I think either is fine.

These would be my leanings for a production environment. Remember that "darko" has production experience also.

Higher-end RAID cards- Really can't beat them. But if I go that way, I keep two similar cards handly or a fallback server that was clustered. Reason? (PRO) If you have a controller failure, on high-end cards you can swap cards and copy back the RAID config and away you go. If have a drive failure within your logical drive, it will pause for a moment, ask what you want to do, then if not caught, time-out, mark the drive as a fault, continue, while rebuilding your array with a spare (if you set up with a spare...). But sometimes in rare instances the controller is "too smart" about false errors. Fault tolerance is settable. Work to do all that functionality is negligible. Price is a consideration- $500 -$2000 for new. You need drivers to see it. Although for myself, I get used or refurbished which the cost is 50-70% off.

Most of the test servers I have at home (10 of them) came with hardware RAID. For most of them, they are a no-brainer. Boot into the on-board BIOS, set them up and go, depending on what OS I am running. Funning that my highest end server is the most PITA to set-up, but that is just an IBM thing. Hardware RAID is well supported by MS Sever 20xx. Most are well supported by Linux, but you have to have a compiling experience/skillset to do it.

Software RAID gives you the flexibility and adaptability of going RAID without the cost or the technology. Think that you can go RAID0, 1, 10, 5, 50, 6, 60... without any extra hardware. Cost is free, Experience and skills in setting up is higher. MDADM is available to install on just about any Linux distro. Once setup, that array can go on and be reused on any box of any hardware config, even on a low-end PC with a Pentium or Athlon as a backup.. set MDADM, read the drives and away you go again. Lower-end cards don't have functionality to shrink, add, grow an array or change RAID types for a migration. (Software RAID does without having to add anything).

Concerns. Drives still fail on either. Functionality is already built into a card, but that functionality is static, meaning what is there is all it's going to have, unless the vendor adds modules or adds something into the firmware. You can add functionality to madadm as you go on, via scripting and your imagination... this again depends on experience and building on your skill-set. You don't need any special specific hardware drivers beyond the basic mdadm package. You can still go this way by setting your smartarray to jbod and no additional cost.

Software mdadm, a lot of things are commandline and manual. That is both good or bad, depending how you look at it. Sometimes you want the control and the chance to mark a drive out as a fault on your own, instead of automatically. Sometimes, you don't want to start degraded... But having things do things automatically are also a blessing if you aren't there physically to get around those things when then happen. The only way to get around those kinds of things when you aren't physically there, or to add functionality on alerts and such, is through scripting.

I know that really didn't answer that "for" you... but it should give you enough to investigate and mold to how "you" do business.

[MAFoElffen]

If I remember it corectly (reading many threads here), you are using 3TB disks with gpt table which means you also needs a small 1MiB UNFORMATTED partition with the bios_grub flag. Grub2 can't install on gpt disk without it. I guess that's the reason it's failing.

See one of my previous posts for the suggested disk layout/partitions. I think it was in this thread.

+1

I even do that for desktops... Translation, I leave 1MB unalloctaed ahead of the first partition. So does "darko."