Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Hi all, First of all: I was very surprised by the degree of sophistication of drbd, a big thanks to the developers! Back to my problem: I successfully managed to set up two (Debian Etch) Xen servers running several Xen guests. Every guest depends on an drbd-device (lets say: /dev/drbd0) for its root filesystem and uses a simple local logical volume for it's swap. The two Xen servers share their drbd-devices, effectively creating a redundant set up of virtual machines. So far, so good. Of course, every drbd-device can only have one primary Xen server (I'm still using drbd 0.7) and the secondary Xen server refuses writing to /dev/drbd0. The primary Xen server runs the actual guest host and allows r/w access to /dev/drbd0. As documented, it's impossible to mount the /dev/drbd0 image on the secondary host, but I found out that it actually is possible to start the Xen guest host using /dev/drbd0. Check out the output of 'xm dry-run mymachine.cfg': > Using config file "mymachine.cfg". > Checking domain: > mymachine: PERMITTED > Checking resources: > phy:/dev/drbd0: PERMITTED > phy:/dev/vg_xen/mymachine-swap: PERMITTED > Dry Run: PASSED Clearly xm doesn't check if it can open /dev/drbd0 for writing and thus allows the user to start the guest host. Of course, this host will hang when trying to write to it's root filesystem, since writing to /dev/drbd0 is not allowed. This seems to be more a Xen bug than a bug in drbd. However, maybe one of you knows a solution for this behavior? I would be perfectly happy if drbd also refused reading from /dev/drbd0 if in secondary mode, but I didn't found out how to configure drbd to act like this. Your suggestions are highly appreciated! Regards, -- Bas van Schaik