Carl’s Blog

This is the second time I have encountered this now. Both sites were completely unrelated and had different IT admins. In each case, for whatever reason EAS was not working, probably due to SSL problem or IP address restrictions, but in trying to fix the problem, the IT person left the machine with ExchangeVdir set to /exchange-oma, but no such virtual directory in IIS.
To fix, I simply removed the ExchangeVdir registry entry, but according to MS KB article 817379, on SBS 2003 the key should indeed be set, and a virtual directory should exist called exchange-oma. Therefore, it seems that the IT person is doing the re-build of Exchange IIS virtual directories (as per the well known MS KB article where you do the metabase edit), but the re-build does not create the SBS-only non-standard exchange-oma directory. I am guessing that the CEICW recreates the exchange-oma virtual directory, otherwise it must be manually created as per kb817379.

In any case, this is something to look for when EAS does not work. The IIS logfiles show the requests for /exchange-oma, which does not exist.

The key issue here is that in attempting to fix a problem, the IT person compounds the problem with exactly the same symptoms but a totally different cause. They probably then move on and fix the initial cause of the problem, but it still doesn’t work because they just goofed up the ExchangeVdir stuff.

Trying to do a physical to virtual conversion so that this blog can be put on the VMWare ESXi 4 box, and the little HP box can do Asterisk (VMWare didn’t play well with Asterisk).
After much frustration (need 32-bit libs on Fedora for the Converter to run), I still couldn’t quite fathom out why the converted machine failed to boot (“cannot find root”), then it dawned on me that the VMWare converter was not finding the root filesystem because it was on a linux raid1 mirror partition, which the converter doesn’t support.

Rather than transferring manually, here’s what I did:
fdisk
change partition type of RAID partition from type fd (linux raid autodetect) to 83 (linux ext). This might not be necessary, but it should prevent the kernel from auto-configuring RAID for that partition.
cd /etc
mv mdadm.conf mdadm.conf.old (i.e. remove the mdadm.conf configuration file, but keep as a backup just in case).
mkinitrd /boot/initrd-noraid-`uname -r`.img `uname -r`
(basically we’re doing “mkinitrd /boot/initrd-noraid-2.2.15-el5.img 2.2.15-el5″ if 2.2.15-el5 is your running kernel version.. the uname -r substitutes this for us.)
cd /boot/grub
edit grub.conf and change kernel boot parameter root= to reflect partition without raid, e.g. change from /dev/md0 to /dev/sda2, also change the initrd= line to /boot/initrd-noraid-x.x.x.img

Now the initial ramdisk has no mdadm.conf, and the partition type is no longer set to linux raid autodetect (type fd).
Power down, remove one of the RAID1 disks, and the system should boot and run now off the other disk without RAID.

VMWare converter now works. Job’s a good ‘un.