I have a system since january or slightly before (say oktober) that has been running a raid 10 on some SB700 motherboard.
Previously I had asked whether it would be possible to define more than 2 arrays (slices) on the same set of disks. AMD tech support told me it was not possible. I still don't know whether it was a limitation of the chipset, or the BIOS. I don't think I ever got a clear answer, althought it was exactly that question that I asked (chipset limitation or BIOS setup limitation).
Because it seemed a very arbitraty limitation (why support more than 1, but not more than 2?)
Today I removed 2 disks from the RAID 10 and as such it turned into RAID 0 effectively, it became a stripe without mirrors.
Then I add another disk to it and rebuild that one disk from the software in Windows (10). At 99% completion it tries to finalize and then reboots. I get a system in which RaidXpert constantly reboots and I have to remove the disk from the array (erase disk, reboot) and readd it, and the same thing will happen again.
Real nice. The thing that is supposed to be your safety measure, now fails.
Since today suddenly the Radeon R9 380 card in it suddenly starts to flicker intensely when Windows 10 is booting. I have no other system on it to test. Linux cannot boot from the software raid from what I have seen (firmware raid) because apparently Grub cannot put the stripes together or something.
But this flickering started happening after at least 3 failed rebuilds and as such 3 spontaneous reboots in that sense.
I have no clue what I am supposed to do. The only thing I could possibly do is to get that system in a (physical) position where I can actually work with it and then just reinstall Windows (10) and hope it gets fixed. And make a backup and note essential configuration etc. I still didn't solve my ideal of having a RAID controller than can do 3 slices, though.
I mean it is not necessary for this system to be the safest of the safest, the RAID 10 was for speed and ease of use in a sense. It was also meant to create 3 "harddrives" out of 4 -- to combine 4 disks in to a RAID and to create 3 "slices" out of that. But I never succeeded in that because:
- AMD doesn't support it
- apparently Intel does
- all of the PCI and PCI-e x1 cards that I have seen do not support it
- I have asked the question of at least 2 suppliers, one of which was a German one that offered both Sil3114 and Marvell 88SX7042 (I think) that told me that none of their cards could do that (have more than one slice).
This would imply that others (a Startech card using Marvell 88SE9230) would probably also not be able to do it. Apart from that, it is PCIe*2 and I don't have one available. I must go with either x1 or PCI (66 MHz) if I go this route. I was just on the phone with Startech (Canada, actually) and they were just such nice people. I know the two-port Adaptec 1210SA cards support what I want, but they are... two port. I have asked a question to Addonics for their ADSA4R5 which comes at a costly 77 euro. Whether it supports it, but the quality (including of the documentation) is not anywhere near that of the Adaptec card. My intent was really just to achieve it with the AMD board.
But AMD tech support was rather inaccurate in their replies and misunderstood my question, and then stated that my motherboard was not even capable of having more than 1 array (which it is). I was inquiring into technical specifics (development potential) and was really hoping to get into contact with a developer himself (or herself).
Today I have a rebooting system that cannot complete the sync to the next disk I insert.
Meaning, I can absolutely not expand my array into the 3rd disk (let alone a 4th).
My questions are:
- is this familiar to anyone? Has anyone seen this behaviour before?
- would you consider it a wise investment to pay more than 75 euro for this addon card, considering that my AMD motherboard MIGHT be able to do it (in firmware, if not in software) given the right BIOS update or the right RaidXpert software?
- Is it sensible to keep using this AMD product seeing as how unreliable it is?
- Is there any potential in trying to reach any AMD developer to see if the thing would actually be capable of being realized in any way?
- Do you suppose there are any hard limits here, or maybe the software could support it easily with just a few minor modifications?
- The software clearly supports creating more than 2 arrays (if on different disks) so the BIOS also supports displaying more than 2 arrays in the boot selection dialog of the regular BIOS. What is going to be the real problem in having more than 2 arrays of the same kind on the same disks? You might even consider that if you changed a single "jump" in the software or a single comparison, it might already support this!!?. Depending on how the list is coded; if there is no list but only a "number1" and "number2" variable, then it might be a lot harder.
- It seems there is no technical limitation unless the chipset itself would not be able to schedule IO calls over more than 2 arrays on the same disks.