Confusion between 'old' and 'new' Raid
It seem to me that many - if not most - of the problems reported by new subscribers to this N.G., are related to a lack of understanding that there are two flavours of Raid, and that the Raid support they have in their stock kernels and the tools supplied with their distribution, relate to the 'old' (obselete) Raid setup. This was the position I found myself in, and only when I read Jakob Ostergaard's excellent Howto and abandoned 'old' Raid, was I able to make any satisfactory progress. Would I not be possible for new subscribers to the mailing list to be automaticallyappraised of the old/new Raid situation via the 'subscribed' response from the list majordomo? Regards: Jim Ford
Large files 2GB+ RAID?
We've been working on getting a x86 Linux system together that would support both RAID and larger then 2 GB file sizes - so far with little luck. RAID works fine on 2.2 kernels, and 2GB files works on 2.3 kernels but RAID doesn't seem to like 2.3 (or at least the 2.3.34 w/ the included RAID). When we attempt to make the ext2 filesystem on the RAID 1 drive, it kernel panics. Seems like this is pretty key for Linux to be used in large data server environments. Has anyone else gotten RAID working with large files on Intel? Thanks for any help / information, Jason. [EMAIL PROTECTED]
Re: Large files 2GB+ RAID?
Unfortunately the hardware RAID still doesn't solve the 2GB+ problem. I also have a hard time with the 'if you want big files, buy a 64 bit machine' argument. What percentage of Linux users are on 64 bit platforms? How many other x86 OS's support 64 bit filesystems (NT, FreeBSD, BeOS, Solaris, etc)? This is a serious impediment to Linux being a server OS, and is most likely going to make us switch to FreeBSD for our projects - even though I much prefer Linux (and know it better). Sure would be nicer to stay with the penguin then turn to the devil Jason. [EMAIL PROTECTED] -- From: Marc Mutz [EMAIL PROTECTED] To: Stephen Waters [EMAIL PROTECTED] Cc: Jason Titus [EMAIL PROTECTED], [EMAIL PROTECTED] Subject: Re: Large files 2GB+ RAID? Date: Tue, Dec 28, 1999, 5:37 PM Stephen Waters wrote: snip 2.3.x are _not_ the kernels one wants to use in production environment. 2.4 will surely have an up-to-date raid implementation. if you need 2GB filesize support and RAID _now_, go buy a non-intel system (alpha, sparc64), or a hardware raid controller that is supported under 2.2.13. that's nice in theory, but a lot of commercial giant database people (oracle, sybase, etc.) only support Linux on x86. So what's wrong with buying a harware raid controller or waiting for 2.4? Marc -- Marc Mutz [EMAIL PROTECTED]http://marc.mutz.com/Encryption-HOWTO/ University of Bielefeld, Dep. of Mathematics / Dep. of Physics PGP-keyID's: 0xd46ce9ab (RSA), 0x7ae55b9e (DSS/DH)
Re: Large files 2GB+ RAID?
Unfortunately the hardware RAID still doesn't solve the 2GB+ problem. I also have a hard time with the 'if you want big files, buy a 64 bit machine' argument. What percentage of Linux users are on 64 bit platforms? How many other x86 OS's support 64 bit filesystems (NT, FreeBSD, BeOS, Solaris, etc)? This is a serious impediment to Linux being a server OS, and is most likely going to make us switch to FreeBSD for our projects - even though I much prefer Linux (and know it better). Sure would be nicer to stay with the penguin then turn to the devil frightenlingly bad puns should be left out of this mailinglist. :) I thought there were motions to backport the BIGMEM patches to 2.2 (unofficial patch of course) any clarification would be useful. -sv
dac960 and weird problem
I have a accelraid 250 w/32mb of ram. I've setup 3 ibm 18 lzx drives (18gig 10krpm LVD drives) on it in a raid 5 configuration. Everything comes up great and functions just fine - but: If I soft reboot the system (ie:ctrl-alt-del or init 6) the dac960 will fail to detect the drives. If I hard-reboot (power off-power on) then everything is peachy. this is a mild annoyance - I'm open to suggestions. I've talked to mylex tech support and they said "change/remove the terminator to see if thats the problem - I did this. - no other suggestions are forthcoming) As well in raid-5 formation the performance on these drives (read and write) is not so hot. I ran bonnie with a -s of 1500MB and the write performance came in at about 5.5-6MB/s - not wonderful - Read was better about 18MB/s - but I thought it would be MUCH better than that. Suggestions are welcome here too. Thanks -sv
Re: Large files 2GB+ RAID?
Ah, sorry for the puns and any confusion. I am talking about 2GB+ file sizes, not memory. The also proves my point - we now have 4GB memory on 32 bit systems - which is only applicable for a VERY small percentage of Linux users, but not 2GB files on 32 bit systems (once again - even though many other 32 bit OSes have them)... Jason My understanding is that the bigmem patches are FS patches not memory patches - they are inappropriately named perhaps. maybe I'm off. -sv
Re: Large files 2GB+ RAID?
Ah, sorry for the puns and any confusion. I am talking about 2GB+ file sizes, not memory. The also proves my point - we now have 4GB memory on 32 bit systems - which is only applicable for a VERY small percentage of Linux users, but not 2GB files on 32 bit systems (once again - even though many other 32 bit OSes have them)... Jason -- From: Seth Vidal [EMAIL PROTECTED] To: Jason Titus [EMAIL PROTECTED] Cc: Marc Mutz [EMAIL PROTECTED], Stephen Waters [EMAIL PROTECTED], [EMAIL PROTECTED] Subject: Re: Large files 2GB+ RAID? Date: Tue, Dec 28, 1999, 6:09 PM Unfortunately the hardware RAID still doesn't solve the 2GB+ problem. I also have a hard time with the 'if you want big files, buy a 64 bit machine' argument. What percentage of Linux users are on 64 bit platforms? How many other x86 OS's support 64 bit filesystems (NT, FreeBSD, BeOS, Solaris, etc)? This is a serious impediment to Linux being a server OS, and is most likely going to make us switch to FreeBSD for our projects - even though I much prefer Linux (and know it better). Sure would be nicer to stay with the penguin then turn to the devil frightenlingly bad puns should be left out of this mailinglist. :) I thought there were motions to backport the BIGMEM patches to 2.2 (unofficial patch of course) any clarification would be useful. -sv
Re: Large files 2GB+ RAID?
Nope. Bigmem was for 4 GB RAM and such, and has been pretty much replaced by highmem (all culled from the Linux Memory Management mailing list). All of the 2GB file stuff is refereed to mostly as Large File Summit (LFS) not to be confused with Log File System (LFS - no idea what it does. Some sort of journal type thing). Once again, any information about large files under RAID would be much appreciated. The pull of FreeBSD is almost inescapable. ahh but freebsd smp is in sore shape in comparison to linux smp. so there is that point. -sv
Re: Large files 2GB+ RAID?
On Tue, Dec 28, 1999 at 06:42:15PM -0500, Jason Titus wrote: by highmem (all culled from the Linux Memory Management mailing list). All of the 2GB file stuff is refereed to mostly as Large File Summit (LFS) not to be confused with Log File System (LFS - no idea what it does. Some sort of journal type thing). Pardon my (mis)understanding of things, but I just had to butt in here. Isn't the question you're asking 'What's the deal with large files under Linux'? I'm not exactly sure how all the RAID stuff works (which is why I'm subscribed to this list..), but AFAIK RAID is just a block device like any other, including a hard disk - so you don't need any specific 'Large files for RAID' patches. It's only the filesystem (and I guess the C library) which must support it. I'm quite sure you can access 2GB on block devices 8). *Maybe* it's possible that you don't have to run ext2 on Linux/RAID? If, for instance, FFS on Linux supported files 2GB (I have no idea if it does), I guess you could run that as the primary filesystem. You could port over the fsck etc. programs from BSD. Might be slower and possibly less reliable, but it might work. I know there are also Large File Summit patches around, but I haven't investigated them (I think the largest file I've got on my system is like 50meg :). I have no clue about this - maybe a post to linux-kernel would help. Otherwise, if you can guarantee that you'll only have a few large files, you can always partition your hard drive and get the apps to write directly to /dev/sda5 instead of a file on a filesystem. But that's probably not very feasible. Once again, any information about large files under RAID would be much appreciated. The pull of FreeBSD is almost inescapable. You're making FreeBSD sound like a bad thing ;). -- : Andre Pang [EMAIL PROTECTED] - Purruna Pty Ltd - ph# 0411.882299 : : #ozone - http://www.vjolnir.org/ozone/:
Silent failure during attampted RAID1 creation
Reply-To: [EMAIL PROTECTED] Hello Gentlemen! I hope you can help me to solve the following problem, 'coz I am stuck :) My goal is to set up RAID1 (mirroring) on two 20GB UDMA66 drives. I have an Abit BP6 mainboard with HighPoint UDMA66 controller, which works fine. The drives are 2 x 20GB Seagate Barracudas, UDMA66. Unfortunately, mkraid refuses to prepare the devices for me: root@serwer | ~# mkraid --really-force /dev/md8 DESTROYING the contents of /dev/md8 in 5 seconds, Ctrl-C if unsure! handling MD device /dev/md8 analyzing super-block disk 0: /dev/hde9, 15221556kB, raid superblock at 15221440kB disk 1: /dev/hdg9, 15221556kB, raid superblock at 15221440kB mkraid: aborted, see the syslog and /proc/mdstat for potential clues. root@serwer | ~# mkraid --really-force /dev/md7 DESTROYING the contents of /dev/md7 in 5 seconds, Ctrl-C if unsure! handling MD device /dev/md7 analyzing super-block disk 0: /dev/hde8, 2104483kB, raid superblock at 2104384kB disk 1: /dev/hdg8, 2104483kB, raid superblock at 2104384kB mkraid: aborted, see the syslog and /proc/mdstat for potential clues. root@serwer | ~# I don't believe anythings goes logged in the syslog. Here's my /proc/mdstat: root@serwer | ~# cat /proc/mdstat Personalities : [3 raid1] read_ahead not set md0 : inactive md1 : inactive md2 : inactive md3 : inactive md4 : inactive md5 : inactive md6 : inactive md7 : inactive md8 : inactive root@serwer | ~# and the uname: Linux server 2.2.13 #2 SMP Tue Dec 28 21:52:25 CET 1999 i686 unknown (Yes, it really has such an original hostname ;- ) Distro: Mandrake 6.1, raidtools-0.90. (Sorry for a lengthy attachments, but they may be useful) TIA 2 all! Every help appreciated (e-mail only, please) Wojtek -- *-- Wojciech Kus ---*--- [EMAIL PROTECTED] -*
Silent failure during attampted RAID1 creation - more
re-hi! (this is a small addition to my prev. mail) I forgot... /etc/raidtab - isn't it useful to know? ;) Here it goes: raiddev /dev/md8 raid-level 1 nr-raid-disks 2 nr-spare-disks 0 chunk-size 4 persistent-superblock 1 device /dev/hde9 raid-disk 0 device /dev/hdg9 raid-disk 1 raiddev /dev/md7 raid-level 1 nr-raid-disks 2 nr-spare-disks 0 chunk-size 4 persistent-superblock 1 device /dev/hde8 raid-disk 0 device /dev/hdg8 raid-disk 1 Thanx, Wojtek -- *-- Wojciech Kus ---*--- [EMAIL PROTECTED] -*
Re: Large files 2GB+ RAID?
[I left off CC'ing the universe on this one...] On Wed, 29 Dec 1999, Andre Pang wrote: On Tue, Dec 28, 1999 at 06:42:15PM -0500, Jason Titus wrote: by highmem (all culled from the Linux Memory Management mailing list). All of the 2GB file stuff is refereed to mostly as Large File Summit (LFS) not to be confused with Log File System (LFS - no idea what it does. Some sort of journal type thing). Logging file systems are a (hard simplification) kind of journaling. Basically, nobody is going to use a true logging filesystem these days - the databases themselves do rollback. The other advantage to logging is the part that is also a part of journaling, and journaling is thus more general purpose. [Pleading for mercy, I'm not a heavy DB guy] *Maybe* it's possible that you don't have to run ext2 on Linux/RAID? If, for instance, FFS on Linux supported files 2GB (I have no idea if it does), I guess you could run that as the primary filesystem. You could port over the fsck etc. programs from BSD. Might be slower and possibly less reliable, but it might work. Large files have to be supported in 3 places: kernel, libc, _and_ the FS. I know there are also Large File Summit patches around, but I haven't investigated them (I think the largest file I've got on my system is like 50meg :). I have no clue about this - maybe a post to linux-kernel would help. My take from the kernel list is that the new style raid stuff shipping on redhat and suse is rock solid in 2.2, but 2.3 is still shaky. There is not that I know of a firm commitment to have new raid in the 2.4 kernel. :( So, look for LFS back-ported to 2.2. Don't run production systems with 2.3 right now. Dark Evil awaits. If anybody will have this soon, it'll probably be Suse due to their LVM work. (Got enough TLA's yet?). Otherwise, if you can guarantee that you'll only have a few large files, you can always partition your hard drive and get the apps to write directly to /dev/sda5 instead of a file on a filesystem. But that's probably not very feasible. Flee in terror. Once again, any information about large files under RAID would be much appreciated. The pull of FreeBSD is almost inescapable. You're making FreeBSD sound like a bad thing ;). Which they are not - and I'm a linux person. :) -- Hunter Matthews Unix / Network Administrator Office: BioSci 222/244 Bioscience Never take candy from strangers. Especially on the Internet.
Re: Silent failure during attampted RAID1 creation
On Wed, 29 Dec 1999, [ISO-8859-2] Wojciech Ku¶ wrote: raiddev /dev/md8 raid-level 1 nr-raid-disks 2 nr-spare-disks 0 chunk-size 4 Do you really want 4-byte chunks? Maybe you meant to use "4k" instead? According to the man page, this value is in bytes. I always add the "k" so as to be explicit. -Andy
resizing raid partitions
Im trying to resize (shrink) my raid partition, ive ran ext2resize and shrank it from 35GB to 15GB, now im not sure how to go about shrinking my /dev/md0 I know the answer has been posted here before (Jakob i think) i couldnt find an archive of this mailing list, does anyone know of any ? I tried raidreconf but i got some error about how it couldnt work with raid type -1 (cant remember the exact message right now, can reproduce it its important) I thought that i could just modify my raidtab and run mkraid -f, but thats my last option as it will probably destroy my data if it fails. Thanks Glenn McGrath
