I was just looking over the patch on blkif.py and I believe there is a bug.
--- a/tools/python/xen/util/blkif.py Fri Dec 10 18:08:19 2010 +0000
+++ b/tools/python/xen/util/blkif.py Wed Jan 05 23:31:24 2011 +0000
@@ -71,15 +71,8 @@ def _parse_uname(uname):
if uname.find(":") != -1:
(typ, fn) = uname.split(":", 1)
- if typ == "phy" and not fn.startswith("/"):
+ if typ in ("phy", "drbd") and not fn.startswith("/"):
fn = "/dev/%s" %(fn,)
- if typ == "drbd":
- if not fn.startswith("drbd"):
- (drbdadmstdin, drbdadmstdout) =
os.popen2(["/sbin/drbdadm", "sh-dev", fn])
- fn = drbdadmstdout.readline().strip()
- fn = "/dev/%s" %(fn,)
if typ in ("tap", "tap2"):
(taptype, fn) = fn.split(":", 1
When you specify a drbd disk for a domU, its format is
wherein drbd expects a resource file in /etc/drbd.d/resourceName.res, that
contains the actual drbd device name (/dev/drbd1 or something).
drbd does not export resources as devices in the /dev/ path (as is
in the above patch). It actually exports them under /dev/drbd/by-res/.
the case in my Debian box, with drbd 8.3.9
I can send out a patch if you folks agree with this.
On Wed, Jan 5, 2011 at 3:32 PM, Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx> wrote:
> Jim Fehlig writes ("Re: [Xen-devel] [PATCH] xend: drbd improvements"):
> > Sorry for the delay here, but I finally got around to investigating this
> > further. As it turns out, drbd's block-drbd script handles all of the
> > details that c/s 20158 introduces within xend :-(. IMO, this c/s should
> > be reverted as it causes a regression. I've tested drbd without this
> > changeset and it works fine.
> Thanks for looking into this and testing things. I have reverted
> 20158 from xen-unstable.
> Xen-devel mailing list
perception is but an offspring of its own self
Xen-devel mailing list