Hi,
I have a RAID-1 configuration looking like this:
#
raiddev /dev/md4
raid-level 1
nr-raid-disks 3
chunk-size 32
nr-spare-disks 0
persistent-superblock 1
device /dev/hdc9
raid-disk 0
device /dev/hda9
raid-disk 1
device /dev/nda
raid-disk 2
#
The "/dev/nda" is a network block device available over
a private network.
The machine running this config is a master into a
fail-over configuration. The other, slave, provide
an own local partition over the net as "/dev/nda" for the master.
The problem I have is when the master is dead (from some reasons)
and the slave is mounting its local partition and starts dealing
with it, the last one (slave machine) didn't know (I don't know)
how to update the event counter to the next value need it
for the reconstruction (on master) when the master is alive.
I want to transfer, some how, the reconstruction only on the master,
such as to tell the master to reconstruct its local partitions,
/dev/hda9 and /dev/hdc9, from /dev/nda and only. For this,
I suppose I need an event counter for /dev/nda grater then
the events stored on /dev/hda9 and /dev/hdc9.
I got two ways:
- a brutal alteration of information stored on /dev/hda9 and /dev/hdc9,
starting from the offset the raid software is reporting on reconstruction
(I saw, that once the RAID is made, this offset is never changed for a partition)
and after start the RAID for this metadisk (/dev/md4). This alteration
is made using " dd if=/dev/zero bs=1024 seek=<offset> count=<total-offset> of=/dev/hdX
"
- or, one solution which I think is more elegant, because it's simple:
create a link on the slave named /dev/nda to the partition exported over the net,
in raidtab.conf declare /dev/hda9 and /dev/hdc9 as failed and after this start
a RAID configuration locally. In this case, the RAID software detect that the name
of the device /dev/nda was changed and will automatically will make the change
on the super block area (or some where else).
Probably will wonder if I have two solutions solving the problem above why
I'm putting this message on the list ? Think about I need to refresh
a disk on a non RAID system (put extra information on it)
and I also need to reconstruct the RAID system (on different machine) from this disk.
I saw if I'll declare on the master machine the local partitions as failed
(of course, changing the order for them, with the /dev/nda as raid-disk 0)
the software start with all three partitions "alive", like nothing was changed,
disregard the failed-disk for /dev/hda9 and /dev/hdc9. And the reconstruction
from /dev/nda never happen.
I'm using 2.4.2 kernel and raidtools-dangerous-0.90-20000116 on RedHat 6.2
Thank you in advance for any points,
--
Adrian Turcu
System Administrator
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to [EMAIL PROTECTED]
