By mistake I installed my raid1 system to use hdc3 as disk 0 and hda3 as
disk 1.
Why is that a `mistake' ?
Now I'm planning to change the disks to another machine that cannot boot from
any other device than hda which means that the current installation would not
work since I'm booting with
What I wanna do is add a spare disk/partition to an existing raid1 array.
Do you want to keep the N-way mirror with N copies, but add a spare so
that when one of the N fails, it will start re-constructing immediately,
or do you want to change it to an N+1 array ?
[ Note that the former will be
Do you want to keep the N-way mirror with N copies, but add a spare so
that when one of the N fails, it will start re-constructing immediately,
...
If the former, just `raidhotadd'.
That is the one I was thinking of using.
Should be straightforward ...
Will it automatically recognize it as
Is there a limit how many md devices I can create?
Yes :-((
My /dev contains md0 to md15. I'd like to leave some gaps between each
device to be able to insert another later (to steer the resync order).
Do you know something I don't ?
Are you saying that the re-sync order is based on the
1. Bring the system down.
2. Install the spare drive
Coward :-)
3. Boot up and edit the /etc/raidtab to include the spare drive
That is not needed, but is good practice.
[[ What I do is to dump the PSB info to /etc/raidtab every night ]]
4. fdisk the drive to be identical to the mirror'd
With RAID1 there is little point in adding a "spare disk" - you can just
add another disk to the running array, ie:
You cannot "add" a third to to a two-way RAID1.
You can create a new three-way RAID1 if you want.
But you can't (AFAIK) change the number of disk is the base array.
md0: sda1
If I do a lilo with boot=/dev/md0 in my lilo.conf it does write to both
disks. But booting from the second disk in the array is not possible.
I always get "error: 0x80" or similar right after the LILO prompt.
Any hints?
I find that `default=linux' (or whatever the actual default is)
and
running on 2.2.14 with raid soft, I have always superblock even after
fsck.ext2.
I assume you mean `I always have a problem with the EXT2 super block' ..
So I guest something was made badly what I can not find out it :(
My guess is that you made an ext2 FS on a partition, then used it under
I am guessing that my problem are the files /etc/md0 and /etc/md1 which
I suspect you mean /dev/md0 and /dev/md1 ...
disappeared when I lost my /dev directory. Is there any way to recreate
these short of rebuilding and therefor destroying the underlining data?
Unless I am missing
1.I made a RAID-0 array (raidtab mkraid /dev/md0) under Mandrake
7.0-2 using a fresh 2.2.14 kernel with IDE RAID patches.
NB: RAID0 and linear I would not expect to do anything.
RAID1, RAID4 and RAID5 are likely to scribble over some of the disks
[ e.g. RAID1 will copy one disk
on a server one disk had a "medium error" and the RAID1 (2.2.14-B1)
disabled one of the mirrors. It looks like this:
md0 : active raid1 sdb5[1] sda5[0](F) 4739072 blocks [2/1] [_U]
We go a JBoD with 24 disks, and about a quarter of them have failed in just
such a way
If I reboot now,
I can make a filesystem and mount it just fine but when I reboot the
kernel does not automatically detect it. I made all of the relevant
partitions type FD and enabled the persistent-superblock. Still no
detection. Any hints?
Do you have the raid0 driver in your initrd ?
I'm running Red Hat 6.1 and I'm getting the following stuff on bootup when
it tries to autodetect my software raid setup:
Do you have an initrd which pre-loads raid5 ?
kmod: failed to exec /sbin/modprobe -s -k md-personality-4, errno = 2
What is wrong?
It needs `md-personality-4' to
I have a related idea. If all of the information in a raidtab can be
derived from a persistent superblock, there should be a tool that allows
generating an actual raidtab file by reading a persistent superblock.
This would be especially useful for dianostic procedures.
I have a perl script
ext2resize claims to be able resize ext2 partitions w/o destroying data.
What's more, it does ...
While there is evidence of this on normal drives and hw raid drives too.
(I assume the `While' is spurious).
I have first hand evidence of the first.
I'd like to know if it will work on sw
What you *REALLY* want is LVM
url please?
Sorry -- I assumed RAID users would all know about http://linux.msede.com/lvm/
Mirror sites
The following ftp sites are known to mirror the LVM tree:
ftp://ftp.gwdg.de/pub/linux/misc/lvm/
(daily mirror, thanks
I've got a four disk RAID5 setup on one controller. I want to add
another controller, but am unsure of what strategy I should adopt to
maintain the RAID integrity.
You didn't mention which raid kernel code you are using.
If it is recent, each partition will have `Persistent Super Block's
in between I learned how to setup a raid1 from existing data. Now I have
one more question: If I decide some fine day that I need one disk of my
raid1 for something different that the mirror, how can I destroy the
raid1 and instead use one of the (old raid) disks as usual (non raid) disk.
I've got two different ide-disks in my system:
/dev/hda -- 6,4GB SAMSUNG SV0644A 11.90 MB/sec
/dev/hdb -- 13.2GB WDC WD136BA 22.22 MB/sec
You aren't going to get on all that well !
Now my question, is it actual to make nearly same sized partitions to use
with RAID-1,
Missing
I've got two different ide-disks in my system:
/dev/hda -- 6,4GB SAMSUNG SV0644A11.90 MB/sec
/dev/hdb -- 13.2GB WDC WD136BA22.22 MB/sec
First i would suggest that you put each disk to separate
channel. hdb -- hdc.
Absolutely and definately 100%, do that.
It thought it, got
for a whole drive:dd if=/dev/zero of=/dev/hdc bs=8k
For a partition: dd if=/dev/zero of=/dev/hdc2 bs=8k
That is indeed an effective way to so it (if somewhat dangerous!)
If it's just to stop them being ext2 FSs, you need not write quite that much !
For ZAPping old PSBs, my `sb.pl'
Q1) Is a `Medium Error' almost certainly a HDA HW problem ?
Q2) Should SCSI errors go to /var/log/messages as well as dmesg ?
Q3) Could RAID problems be reported by email ?
Q4) How `should' iffy disks be replaced ?
Running RedHat 6.1/2.2.12-20smp on a Gigabyte GA-6BXDU (Adaptec 7896),
to access
It could be handy if there was another 'file' in /proc/sys/dev/md that
could dump an equivalent raidtab of the running arrays. Or perhaps a
separate raidtab for each array: raidtab.md0 raidtab.md1, etc... Is this
possible? Should I be the one to get gung-ho about it if I want this bad
What does this means ?
It would thrash the disk heads to keep moving from one partition to another
between each disk operation, so when re-synching disks, it ensures that only
one thread is using a particular disk (spindle) at a time.
Where is the overlapping ?
Not really overlapping,
Exec summary: will raid1 code cause read error sectors to be re-vectored ?
Full gory details:
I am having problems with some new disks, in that sectors are going bad.
I kind of didn't expect this on `new' disks, but maybe I'm just not used to
having so many sectors !
So far, the problems
Now, before someone yells at me for being off topic :), does anyone know
how the Linux software RAID handles it if a new drive was inserted in the
SCSI system in a device letter preceding a RAID configured for
autodetection?
yes -- each SB holds information on `this' partition as well as all
How easy is it to move a RAID device?
Trivial -- shut down the system, move the device, reboot !
Specifically, I have six SCSI drives on the same controller which I want to
split between two controllers.
With auto-detection of RAID devices, this comes for free !
[[ With SCSI this is fairly
kmod: failed to exec /sbin/modprobe -s -k md-personality-3, errno = 2
...
has anyone seen a similar problem?
No, but it seems to be saying to me that modprobe can't find raid1.
Try adding the line
alias md-personality-3 raid1
to /etc/conf.modules (or whatever you have) and see if it
1. How do I mirror /usr without destroying the data on it.
Assuming /usr is a partition, yes. I am in the process of writing a page of
private suggestions -- http://www.cl.cam.ac.uk/~pb/linuxraid-faq.html
Forgive my novice question, but is this implying that with a SCSI
arrangement, failure of one drive will not lock/affect the SCSI bus?
Failure of a **DRIVE** should not effect the SCSI bus at all.
Failure of the *controller* (real failures are very rare, normally it is power
loss) can cause
after purposely trying to corrupt one drive in my raid1 array (for testing
purposes of course),
RAID1 does not protect against disk corruption.
It protects against disk *FAILURE*.
i setup raid1 array w/ 2 identical 9gig scsi drives, and made an ext2fs on the
array. then i stopped the
During the init scripts, it does try to start and mount the RAID, but fails.
My guess is that you have raid0 as a loadable module -- is that so ?
You can check using (on a machine with an active md) something like:
% grep raid /proc/modules
raid1 6080 1
who got experience or an "estimate for results" for the following plan:
Sorry -- only the latter ... :-(
I have used Partition Magic 4.0 several times on non-raid systems.
I didn't encounter any trouble shrinking, expanding, creating or moving
ext2fs Partitions.
Did you ever move a primary
We've experienced a few odd anomalies during testing our IDE RAID-5
array ( 6 x 16gb =80g).
raidtools 0.90 with raid0145 patches ?
We simulated a drive failure by disconnecting a drive's power, and if
the IDE channel contained a second drive in the RAID5 array, the array
was permanently
I've finally decided to bite the bullet and try to get devfs working (so that
I can have lots of small (1GB) partitions on my SCSI disks (as LVM doesn't
seem to be on the horizon yet).
Unfortunately, the raid0145 and devfs patches clash.
I thought that it might be possible to remove the raid
(I'm aware of the failed-disk patches and have used them
successfully. Is there another way?)
(resize2fs from PartitopnMagic, and of course dump/restore ...)
If you ask me, mke2fs should be modified to reserve that 4k in the end
of the volume by default..
Yup -- I too suggested that when I
I tested this some time ago ( dd if=/dev/hda1 of=/dev/hdc1 in single user
mode) but did not shrink the ext2fs, but it worked.
That is not a surprise -- the PSB is onlt one 4K block at the very end ...
Will the superblock get corrupted when I use up my last 4k of space on the
ext2fs ?
I've asked before that the AUTODETECT code always look for PSBs in generated
RAID partitions, so that recursive RAID partitions work (e.g. RAID1 10/0+1/1+0)
Does anyone know if this has been done ?
I have started them manually and it seems to work.
Do I just have to HACK it by calling "raidstart
I am making several tests but I haven't found information about how I
can replace a broken disk.
raidhotadd
- Then raid system boots in degraded mode and don't start the
reconstruction process.
raidhotremove /dev/md0 /dev/hdd2# just to be tidy ...
raidhotadd
[root@news /root]# /sbin/raidstart -a
/dev/md0: Invalid argument
As ever, look in /var/log/messages ...
I suspect you should be using "mkraid /dev/md0" ...
May 9 16:19:07 news kernel: md: invalid raid superblock magic on sdb1
Not surprising -- there is no SB ...
I suspect you should be using "mkraid /dev/md0" ...
But if I did mkraid /dev/md0 wouldn't that destroy all the data on the array?
Not under 0.90 as fasr as I know -- as has been
so what shoud I run? mkraid /dev/md0 ?
That is my impression.
or mkraid /dev/md0 --force-resync -f ??
With linear, there is no "sync" ...
what do I need to do to access my data?
My impression is that you need:
mkraid /dev/md0
mount /dev/md0
[ It's great when you can pass on knowledge and other "see the light" and
pass it on in turn :-))
]
Whereas Tom has does it "for real" with his many GB of data depending on it, I
have never run a RAID5 array (other than 3 cylinders on a single disk!) so am
free to postulate from my ivory
Running RedHat Starbuck 5.9.7, kernel 2.2.6, mkraid version 0.90.0
from the raidtools-0.90-3 RPM package.
This looks like the FAQ:
Q) Why does raidtools-0.90 fail on my standard 2.2.[56] kermel
A) raidtools-0.90 requires raid145 patches to be applied to the kernel
mkraid: aborted
That
Any chance of you providing a 'cookbook' on howto do this? Just an
illustration of the commands to execute and what to do in what order?
Note that for the fancy / clean way of doing this (rather than just ZAPPing
the partition type, the SB or some such) needs patches to the patches ...
Also,
both Dell 2300's with RedHat 5.2 installed. I have upgraded the kernel of
this machine to 2.2.5, including RAID 0 and RAID1 support in that kernel. I
downloaded the 3/7/99 version of Raidtools .90 and installed them.
Have you applied the "raid145" patch to the kernel ?
When I try to mkraid,
I'm a fan of SW RAID (a motherboard with 3 LVD busses, 2 EIDE busses, and 1 W
bus kind of satisfies our disk IO requirements), but I've being pressured to
buy some HW RAID for our "main" NFS servers.
The suggestion is either AMI 438 (which I was suprised not to see mentioned in
i have feel like that i am missing something obvious :-)
What does /var/log/messages say ?
Could the information be stuck in the master boot record?
Any LILO gurus out there who know about "spare" space in the MBR ??
Anyhow, as far as I can see the real issue is how to replace a failed drive
in the array..
No -- that's easy -- raidhotadd ...
In a RAID setup, you'd like to be
That's not the point: we don't even _look_ for a raid superblock unless
the partition is marked for autostart.
Yup -- a *REAL* bummer if the partition doesn't have a partition type.
There are related problems we need to deal with regularly when building
filesystems: what happens if you
you cant autodetect with modules. must compile raid into kernel.
Not true -- using initrd modules work ...
Sorry I forgot to say I compiled raid support for 0 and 1 into the
kernel, and 5 as a module. But I don't use it
Then you'll have to give more info, e.g. what the RAID code said
My problem is that whenever I boot now, the array comes
up without sda2. I can raidhotadd sda2 and things are happy
until the next reboot.
Hmm -- sounds to me like "raidhotadd" isn't re-writing the SB -- sorry, I don't
know how that code works :-(
What does a "mkraid --debug" give you in
As there was no immediate interest im a kosher "superblock" command,
I HACKed together a perl script to do the business (save SB to file,
write SB from file, set datestamps, manipulate disk states, etc) which
does indeed include the ability to generate /etc/raidtab entries ...
That would be
Are there any plans to have a utility as in Solaris (Disksuite) where
metastat can output the config format so that the equivalent of raidtab
can be updated...
That was indeed one of the suggestions I made ...
As there was no immediate interest im a kosher "superblock" command,
I HACKed
Should raidtools-19990309-0.90 manage a linear device without a SB ?
[ I can "mkraid" it, but once stopped, it can never be restarted ]
md8 fails, md7 is fine.
==
raiddev /dev/md7
raid-level linear
[ raidtools-19990309-0.90, Linux 2.2.2 + raid0145-19990309 ]
If I create a single disk mirror, then raidhotadd a second disk, it tries to
resync, but gets nowhere, blocks all other md re-sync's, make the load average
make the machine look busy (it is actually idle), and means that any "sync"
Since it's nonpersistent, it can only be re-created.
Ahh -- which "raidstart" does not do ... right ...
The 'old' mdadd+mdrun was always re-creating arrays as well.
So that has to be used ... [ but cannot share a /etc/raidtab :-( ]
raidstart (and autostart) starts only 'persistent'
Does one have to manually update /etc/raidtab to show which disks are now
"spare" and which are the main disks ?
raidhotadd doesn't consult the raidtab file, all the info it needs is
contained in the raid superblock. Since the disk you add gets used right
away the info in /etc/raidtab should
if you hotadd the (repaired or replaced) /dev/sda1 again you'll have
raid-disk1/dev/sda3
raid-disk2/dev/sda2
spare /dev/sda1
So "Yes, you do have to manually edit /etc/raidtab to keep in sync". Ta.
the patch to raidtools + kernel
Which patch to
The functionality of ckraid is automatically performed by kernel on startup
of raid devices.
Is there a way to say "check the current status in verbose mode" ?
/proc/mdstat shows the "bare bone" info, but not things like spare disks, etc.
raidhotadd /dev/md0 /dev/sda2
Ta -- it's not
[root@xx raidtools-0.90]# mkraid --force /dev/md0
DESTROYING the contents of /dev/md0 in 5 seconds, Ctrl-C if unsure!
handling MD device /dev/md0
analyzing super-block
disk 0: /dev/sdb1, 4188937kB, raid superblock at 4188864kB
disk 1: /dev/sdc1, 4188937kB, raid superblock at 4188864kB
[ I'm new to the list, so feel free to point me at TFM ... ]
I have built a Linux 2.2.2 kernel with raid0145-19990309-2.2.3 and RAIDs
0 and 1 seem to work just fine (linear seems to need a chunksize!).
With the demise of "ckraid", how can I do I "check all is well" test ?
Also, how do I say
62 matches
Mail list logo
