Hi,
I've been running an AoE service at my shop for general storage - nothing major, just a 4TB target mounted on Windows for sharing (everyone at the shop but me are Windows-nerds, I prefer Linux). Everything was fine for about 2 years when my Windows 7 machine had a hardware failure so I installed the initiator on a windows 8 box in the shop - it worked for about a day, and now the target is suddenly no longer visible to windows (the error it reports is corrupt/missing data - not very useful). The Target is on CentOS 6.6 and is a striped LV running on vbladed - it's still up and the volume appears to be intact. Since the volume has always been mounted and managed by a windows AoE initiator, the filesystem is not ExtN... it should be NTFS, but I can't manually mount it as NTFS on my CentOS box.
The Client is running Starwind Initiator build 2010-03-10
The Target (CentOS 6.6 - x64) is running:
aoetools: 36
installed aoe driver: 81
running aoe driver: 81
The Target supports udev (udev config below) - but I guess I just don't understand how the dynamic mapping works. Prior to the failure I was NOT running with udev, but I didn't change anything except installing regular updates with yum. I assume a kernel update somewhere in the recent past added and implemented support for udev and everything got confused???
My udev rules are as follows:
SUBSYSTEM=="aoe", KERNEL=="discover", NAME="etherd/%k", GROUP="disk", MODE="0220"
SUBSYSTEM=="aoe", KERNEL=="err", NAME="etherd/%k", GROUP="disk", MODE="0440"
SUBSYSTEM=="aoe", KERNEL=="interfaces", NAME="etherd/%k", GROUP="disk", MODE="0220"
SUBSYSTEM=="aoe", KERNEL=="revalidate", NAME="etherd/%k", GROUP="disk", MODE="0220"
# aoe block devices
KERNEL=="etherd*", NAME="%k", GROUP="disk"
The files /dev/etherd/discover, err, interfaces, revalidate all exist and the LV target is in a /dev subdirectory symbolically-linked back to the a block device (/dev/dm-5). LVM reports the volume is intact and healthy.
Please let me know if there are any links I should read regarding this - I've searched all over the place and nothing seems to address the difference between udev-dynamic and statically mapped device methods for AoE. I'm not even sure that's the problem.
Also let me know if I need to post any configuration info that might help.
Thanks so much in advance...
Pulling what hair I have left out,
-DM
The Latest Gartner® Magic Quadrant™Hyperconverged Infrastructure Software