Hello,
I tried to do the following:
Computer 1 (node1) | Computer 2 (node2)
-------------------------------------+------------------------------------
dd if=/dev/zero of=/home/nbd (100MB) | dd if=/dev/zero of=/home/nbd (100MB)
nbd-server 1024 /home/nbd | nbd-server 1025 /home/nbd
| nbd-client node1 1024 /dev/nd0
| nbd-client localhost 1025 /dev/nd1
| mkraid -c nbdtest /dev/md1
| mke2fs -b 4096 /dev/md1
| mount /dev/md1 /mnt
Content of nbdtest:
raiddev /dev/md1
raid-level 1
nr-raid-disks 2
nr-spare-disks 0
chunk-size 4
persistent-superblock 1
device /dev/nd1
raid-disk 0
device /dev/nd0
raid-disk 1
Up to this point everything went fine. I now copied files (on node2) via
mc into this file system. After copying about 50 files mc just locked up
and I could see via ps that it was in D state. The load on node2 was 2
but could not see (with top) any process causing this. So it must have
been the kernel. Funny thing is I could still log in on node2, but as
soon as I wanted to create a file, that process would hang up in D state.
I was not able to reboot node2 so I had to press the reset button.
System setup node1: 486DX-133 32MB with one 2GB SCSI disk,
kernel 2.2.14 + mingos patch
node2: 2 x PII-350 128MB with 5 disks used as one single
SW-Raid5, kernel 2.2.14 + mingos patch
node1 and node2 are connected via two ne2000 cards (10Mbit).
The aim of this exercise was to have a file system that can be mounted by
say node1 and if it goes down node2 can mount it and the application on
node2 can continue with the data at that point where node1 crashed.
Has anyone tried this SW-Raid1 over network block devices and is
that what I am trying to do possible?
Is it a problem that /dev/nd1 lies on another SW-Raid? ie. Part of a raid1
on top of a raid5.
Or are there other ways of achieving this?
Holger
