Re: xRAID disks....
On 12 Jun 2008 , DA Forsyth entreated about "Re: xRAID disks": > hmmm, thinking now the gmirror create/remove route will probably > work. let me try it on a blankish disk and see this appears to be the answer to the question: how to stop ar recognizing a disk that used to be on a raid controller doing a 'gmirror label gm0 /dev/ad1' filled the last sector with data, and 'gmirror clear /dev/ad1' reset it all to zero now to try it on the big disk with data on it... YES: it works, and the data slice is still there, and the commands happen a lot faster than a dd with a skip parameter sidenote: to see the last sector use sysinstall's fdisk to see the data for the disk. you'll see something like Disk name: ad3FDISK Geometry: 38913 cyls/255 heads/63 sectors = 625137345 sectors Offset Size(ST)End Name PType Desc Subtype 0 63 62- 12 unused0 63 625137282 625137344ad3s1 8freebsd 165 625137345 5103 625142447- 12 unused0 ^ you want that number in a dd commmand like this dd if=/dev/ad3 skip=625142447 | hd -v Thanks to all for the pointers -- DA Fo rsythNetwork Supervisor Principal Technical Officer -- Institute for Water Research http://www.ru.ac.za/institutes/iwr/ ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: xRAID disks....
On 10 Jun 2008 , Erik Trulsson entreated about "Re: xRAID disks": > On Tue, Jun 10, 2008 at 04:56:07PM +0200, Wojciech Puchar wrote: > > > > > > The pair of ex-RAID disks are ad1 and ad2 and obviously are no longer > > > a raid pair, yet 'something' is telling the ar() driver to try and > > > pair them and failing because there is no raid hardware in that box. > > > > there are no "raid hardware" on most devices. it's just marketing hype. > > Most (cheap) RAID controllers do almost everything in software. Some do have > hardware support. this was not a cheap card, it is an Adaptec 2400A. 4 disks. It can do RAID5 too but I never tried that, having needed 2 mirror pairs instead. the new motherboard is an Intel D965 with 6 SATA sockets, 4 of which are now supporting the new RAID5 array (4x400Gb disks). whether is is the RAID hardware or just because it is SATA2, it is damn fast compared to IDE. pleased so far. oh, not booting off this, it is just data (and the old IDE mirror disks must now work in the non RAID backup server) -- DA Fo rsythNetwork Supervisor Principal Technical Officer -- Institute for Water Research http://www.ru.ac.za/institutes/iwr/ ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: xRAID disks....
On 10 Jun 2008 , Erik Trulsson entreated about "Re: xRAID disks": > > I suspect the raidinfo is stored on the disk somewhere and a suitable > > 'dd' command can erase it. but where and how? > > That kind of information is usually stored last on the disk (where it is > least likely to be overwritten by filesystems, partitioning info, or boot > loaders), so if you overwrite the last couple of KBs on those disks you will > probably be fine. > (If you want to be certain you can always use 'dd' to nuke all the > information on the disk. That will take longer time, but you get the extra > advantage of testing all the blocks on the disk so that they work > correctly.) > > For the first you could do something like: > dd if=/dev/zero of=/dev/ad1 bs=1m skip=76318 > which should overwrite the last MB of ad1 with zeros. I tried to overwrite just the last sector, but though dd reported success (and took ages, seems it has to do a read for every skipped sector) the data was still there when I used dd to display it. I have just done a search for sector editing software but I cannot find anything in ports. Starting to think of writing some C... how hard can it be just to seek to a given sector and scribble zeros on it? so then I did this (overwrite last megabyte)and that did in fact zero the last megabyte, taking away the raid info AND all the partition info. not exactly what I wanted but I was going to repartition anyway. but I now have another disk with data and raid info on it and will need a way to nondestructively remove the raid info there. I did try the suggestion of 'atacontrol' but it did nothing, I also tried 'gmirror clear' but that gives an error message, maybe I should first create a gmirror then clear it. or maybe try the 'forget' command hmmm, thinking now the gmirror create/remove route will probably work. let me try it on a blankish disk and see -- DA Fo rsythNetwork Supervisor Principal Technical Officer -- Institute for Water Research http://www.ru.ac.za/institutes/iwr/ ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: xRAID disks....
You do not normally have that much bandwidth even in a modern machine. Typical bandwidth for the northbridge/southbridge connection is 1-2 GB/s for most machines sold today. (For example just about all machines with a recent Intel desktop chipset. The connection between north- and south-bridge on those is equivalent to a PCI-E x4 connection (which provides 1GB/s in each direction.)) as long as it's not saturated it's not a problem. with several other devices (which is not uncommon) then the reduced bandwidth usage can be very useful. true. but not if it's builtin in chipset or on PCI express. PCI-E controller cards are still fairly uncommon, and many of them but integrated in chipset - common. require a x4 or x8 slot, while most motherboards only have x1 slots (apart from the x16 slot intended for a graphics card.) this slot is usable for anything. i always take some old PCI card for free for servers. as they don't need graphics anyway. on my 8-disk server i could get 95MB/s from EACH of 8 drives in parallel, still having minimal system load. it isn't anything expensive, quite cheap gigabyte motherboard with core2 duo and 2GB RAM ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: xRAID disks....
On Tue, Jun 10, 2008 at 05:13:31PM +0200, Wojciech Puchar wrote: > x> hardware support. > > > >> actually there is not much need to have it for RAID-0/1/10 where there is > >> almostnothing to process. > > > > For mirrors it can actually be a big win with hardware support. > > If you use software RAID then you will have to perform each write twice > > (once to each disk), > in parallel > > while with hardware support for RAID you only need > > to transfer the data once. > > which saves at most 100MB/s bandwidth - compare this to 5-10GB/s in modern > machines. You do not normally have that much bandwidth even in a modern machine. Typical bandwidth for the northbridge/southbridge connection is 1-2 GB/s for most machines sold today. (For example just about all machines with a recent Intel desktop chipset. The connection between north- and south-bridge on those is equivalent to a PCI-E x4 connection (which provides 1GB/s in each direction.)) And that is for modern machines. Older ones have even less bandwidth available. > > If the controller resides on a PCI-bus together > > with several other devices (which is not uncommon) then the reduced > > bandwidth usage can be very useful. > > true. but not if it's builtin in chipset or on PCI express. PCI-E controller cards are still fairly uncommon, and many of them require a x4 or x8 slot, while most motherboards only have x1 slots (apart from the x16 slot intended for a graphics card.) (And PCI-express is still fairly new, so there are lots of computers in use that do not have any PCI-E slots at all.) If you go back just a few years you will find that chipset itself provides only two IDE-channels and nothing more. Any other devices reside on a single PCI-bus (which provides a total bandwidth of 133MB/s.) > > > there are really not worth price. unless you need RAID-5. > > but with todays disk prices it's better to just use RAID-1+0 and bigger > drives. That depends on what your goals are, and what constraints you operate under. RAID 10 is nice, but it requires more disks then RAID5. Extra disks create extra noise and require more power and generate more heat and (most importantly) require extra space. There are a limited amount of space available in most computer cases, which might not be able to accomodate the extra disks needed for RAID10. > > with software RAID you are not forced to operate on whole disks. usually > not everything has to be mirrored. If you have reason to use mirroring at all, then I would say that just about everything should be mirrored. (RAID is in no way a substitute for backups. The main reason for using RAID is either performance (which is often better served by several independent disks anyway) or to minimise downtime. If some parts of your disks are not mirrored then you be able to avoid that downtime anyway.) -- Erik Trulsson [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: xRAID disks....
Erik Trulsson wrote: On Tue, Jun 10, 2008 at 04:33:47PM +0200, DA Forsyth wrote: Hiya I've had this for a while now and have done many searches for info but have not yet come up with the right question, hence have not got the answer. My main server has an Adaptec IDE raid card. A couple of years ago I took disks that had been a mirror pair on that card out of the server and put them into my test server, not as a raid pair since the test server has no raid hardware. During boot I see this ad0: 19092MB at ata0-master UDMA66 ad1: 76319MB at ata0-slave UDMA66 ad2: 76319MB at ata1-master UDMA66 ad3: 19092MB at ata1-slave UDMA66 ar0: 76319MB status: BROKEN ar0: disk0 DOWN no device found for this subdisk ar0: disk1 DOWN no device found for this subdisk The pair of ex-RAID disks are ad1 and ad2 and obviously are no longer a raid pair, yet 'something' is telling the ar() driver to try and pair them and failing because there is no raid hardware in that box. Now I am reconfiguring that machine a bit and would like to fix this, both on these existing drives and on the 320MB drive I have just removed from a RAID1 pair and will be putting into the box instead of ad3 (the other 320GB from the pair is in a USB enclosure for other purposes and has not shown any signs of knowing it was in a raid pair) I suspect the raidinfo is stored on the disk somewhere and a suitable 'dd' command can erase it. but where and how? That kind of information is usually stored last on the disk (where it is least likely to be overwritten by filesystems, partitioning info, or boot loaders), so if you overwrite the last couple of KBs on those disks you will probably be fine. (If you want to be certain you can always use 'dd' to nuke all the information on the disk. That will take longer time, but you get the extra advantage of testing all the blocks on the disk so that they work correctly.) For the first you could do something like: dd if=/dev/zero of=/dev/ad1 bs=1m skip=76318 which should overwrite the last MB of ad1 with zeros. To erase all of the disk: dd if=/dev/zero of=/dev/ad1 bs=1m It's rather easier and quite a lot less risky to simply do: # atacontrol delete ar0 Cheers, Matthew -- Dr Matthew J Seaman MA, D.Phil. 7 Priory Courtyard Flat 3 PGP: http://www.infracaninophile.co.uk/pgpkey Ramsgate Kent, CT11 9PW signature.asc Description: OpenPGP digital signature
Re: xRAID disks....
Wojciech Puchar wrote: I recently removed one pair of disks from a windows "hardware :)" RAID controller, and upon inserting it into a newly built FreeBSD system, it was immediately detected by the ar driver, and messages started coming in, like is there actually any difference in ar and gmirror/gstripe except that ar is simpler, takes only whole drives and use "hardware RAID ;)" compatible headers ? make the disks forget about their previous RAID-life: atacontrol delete ar0 Have a look at man atacontrol. I have not tried it, but it is probably worth a try. I've used gmirror on several occasions, and it works well for me. I have never used ar, but as I understand this is limited to ata disks (hence ar=atapi raid). The geom framework probably provides a lot more features and is not limited the way ar is. I would not mind using ar in this particular system, but I moved disks one by one, erasing the first one (ar was still complaining about the other one missing after I erased it) and when I later added the second one, I got a kernel panic. Mind you, the ar "signature" or whatever was written in a windows system and may not have been exactly compatible. I removed ar from the kernel and continued with my usual gmirror stuff (which works flawlessly) ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: xRAID disks....
I recently removed one pair of disks from a windows "hardware :)" RAID controller, and upon inserting it into a newly built FreeBSD system, it was immediately detected by the ar driver, and messages started coming in, like is there actually any difference in ar and gmirror/gstripe except that ar is simpler, takes only whole drives and use "hardware RAID ;)" compatible headers ? make the disks forget about their previous RAID-life: atacontrol delete ar0 Have a look at man atacontrol. I have not tried it, but it is probably worth a try. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: xRAID disks....
x> hardware support. actually there is not much need to have it for RAID-0/1/10 where there is almostnothing to process. For mirrors it can actually be a big win with hardware support. If you use software RAID then you will have to perform each write twice (once to each disk), in parallel while with hardware support for RAID you only need to transfer the data once. which saves at most 100MB/s bandwidth - compare this to 5-10GB/s in modern machines. If the controller resides on a PCI-bus together with several other devices (which is not uncommon) then the reduced bandwidth usage can be very useful. true. but not if it's builtin in chipset or on PCI express. there are really not worth price. unless you need RAID-5. but with todays disk prices it's better to just use RAID-1+0 and bigger drives. with software RAID you are not forced to operate on whole disks. usually not everything has to be mirrored. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: xRAID disks....
Erik Trulsson wrote: On Tue, Jun 10, 2008 at 04:33:47PM +0200, DA Forsyth wrote: Hiya I've had this for a while now and have done many searches for info but have not yet come up with the right question, hence have not got the answer. My main server has an Adaptec IDE raid card. A couple of years ago I took disks that had been a mirror pair on that card out of the server and put them into my test server, not as a raid pair since the test server has no raid hardware. During boot I see this ad0: 19092MB at ata0-master UDMA66 ad1: 76319MB at ata0-slave UDMA66 ad2: 76319MB at ata1-master UDMA66 ad3: 19092MB at ata1-slave UDMA66 ar0: 76319MB status: BROKEN ar0: disk0 DOWN no device found for this subdisk ar0: disk1 DOWN no device found for this subdisk The pair of ex-RAID disks are ad1 and ad2 and obviously are no longer a raid pair, yet 'something' is telling the ar() driver to try and pair them and failing because there is no raid hardware in that box. Now I am reconfiguring that machine a bit and would like to fix this, both on these existing drives and on the 320MB drive I have just removed from a RAID1 pair and will be putting into the box instead of ad3 (the other 320GB from the pair is in a USB enclosure for other purposes and has not shown any signs of knowing it was in a raid pair) I suspect the raidinfo is stored on the disk somewhere and a suitable 'dd' command can erase it. but where and how? That kind of information is usually stored last on the disk (where it is least likely to be overwritten by filesystems, partitioning info, or boot loaders), so if you overwrite the last couple of KBs on those disks you will probably be fine. (If you want to be certain you can always use 'dd' to nuke all the information on the disk. That will take longer time, but you get the extra advantage of testing all the blocks on the disk so that they work correctly.) For the first you could do something like: dd if=/dev/zero of=/dev/ad1 bs=1m skip=76318 which should overwrite the last MB of ad1 with zeros. To erase all of the disk: dd if=/dev/zero of=/dev/ad1 bs=1m I recently removed one pair of disks from a windows "hardware :)" RAID controller, and upon inserting it into a newly built FreeBSD system, it was immediately detected by the ar driver, and messages started coming in, like in your case. Although I ended up removing the ar device from the kernel (I was going to use gmirror), I found out that there is possibly another way to make the disks forget about their previous RAID-life: atacontrol delete ar0 Have a look at man atacontrol. I have not tried it, but it is probably worth a try. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: xRAID disks....
On Tue, Jun 10, 2008 at 04:56:07PM +0200, Wojciech Puchar wrote: > > > > The pair of ex-RAID disks are ad1 and ad2 and obviously are no longer > > a raid pair, yet 'something' is telling the ar() driver to try and > > pair them and failing because there is no raid hardware in that box. > > there are no "raid hardware" on most devices. it's just marketing hype. Most (cheap) RAID controllers do almost everything in software. Some do have hardware support. > actually there is not much need to have it for RAID-0/1/10 where there is > almostnothing to process. For mirrors it can actually be a big win with hardware support. If you use software RAID then you will have to perform each write twice (once to each disk), while with hardware support for RAID you only need to transfer the data once. If the controller resides on a PCI-bus together with several other devices (which is not uncommon) then the reduced bandwidth usage can be very useful. (And for RAID you will need at least some support on the controller if you want to be able to boot from a striped volume.) > > > purposes and has not shown any signs of knowing it was in a raid > > pair) > > > > I suspect the raidinfo is stored on the disk somewhere and a suitable > > 'dd' command can erase it. but where and how? > erase the whole drive. > > and next time don't use "hardware" RAID anymore. use gmirror and gstripe > to have PORTABLE RAID. -- Erik Trulsson [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: xRAID disks....
On Tue, Jun 10, 2008 at 04:33:47PM +0200, DA Forsyth wrote: > Hiya > > I've had this for a while now and have done many searches for info > but have not yet come up with the right question, hence have not got > the answer. > > My main server has an Adaptec IDE raid card. A couple of years ago I > took disks that had been a mirror pair on that card out of the server > and put them into my test server, not as a raid pair since the test > server has no raid hardware. > > During boot I see this > ad0: 19092MB at ata0-master UDMA66 > ad1: 76319MB at ata0-slave UDMA66 > ad2: 76319MB at ata1-master UDMA66 > ad3: 19092MB at ata1-slave UDMA66 > ar0: 76319MB status: BROKEN > ar0: disk0 DOWN no device found for this subdisk > ar0: disk1 DOWN no device found for this subdisk > > The pair of ex-RAID disks are ad1 and ad2 and obviously are no longer > a raid pair, yet 'something' is telling the ar() driver to try and > pair them and failing because there is no raid hardware in that box. > > Now I am reconfiguring that machine a bit and would like to fix this, > both on these existing drives and on the 320MB drive I have just > removed from a RAID1 pair and will be putting into the box instead of > ad3 (the other 320GB from the pair is in a USB enclosure for other > purposes and has not shown any signs of knowing it was in a raid > pair) > > I suspect the raidinfo is stored on the disk somewhere and a suitable > 'dd' command can erase it. but where and how? That kind of information is usually stored last on the disk (where it is least likely to be overwritten by filesystems, partitioning info, or boot loaders), so if you overwrite the last couple of KBs on those disks you will probably be fine. (If you want to be certain you can always use 'dd' to nuke all the information on the disk. That will take longer time, but you get the extra advantage of testing all the blocks on the disk so that they work correctly.) For the first you could do something like: dd if=/dev/zero of=/dev/ad1 bs=1m skip=76318 which should overwrite the last MB of ad1 with zeros. To erase all of the disk: dd if=/dev/zero of=/dev/ad1 bs=1m -- Erik Trulsson [EMAIL PROTECTED] ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: xRAID disks....
The pair of ex-RAID disks are ad1 and ad2 and obviously are no longer a raid pair, yet 'something' is telling the ar() driver to try and pair them and failing because there is no raid hardware in that box. there are no "raid hardware" on most devices. it's just marketing hype. actually there is not much need to have it for RAID-0/1/10 where there is almostnothing to process. purposes and has not shown any signs of knowing it was in a raid pair) I suspect the raidinfo is stored on the disk somewhere and a suitable 'dd' command can erase it. but where and how? erase the whole drive. and next time don't use "hardware" RAID anymore. use gmirror and gstripe to have PORTABLE RAID. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"