Hi folks,
I''ve been asked to post my experiences here, so... well, here goes :)
(in no particular order).
We can view bugs in kernel/xen via bugs.opensolaris.org, but not file
them against kernel/xen. Could that be added to the list of categories
for submission?
There''s kind of a composite set of problems with bootadm(1M).
While the modified bootadm will update bootadm-created entries, in
menu.lst, it *won''t* modify those created by liveupgrade.
As an extension of that, the bootadm created entries in menu.lst are not
always the *correct* entries to modify in a liveupgrade environment.
(there''s also the issue that the ''real'' menu.lst is
not always on the
currently active BE. I can''t currently work out if that''s
dealt with
naturally, or not, because of the above).
When BFU''ing the xen bits into a liveupgrade BE, the bootadm-created
menu.lst entry will be updated for the new booting scheme. The
liveupgrade-created entries will not. Additionally, the bootadm-created
entries don''t stand much chance of referring to the BE BFU''d
with xen
bits. This leaves you hand-editing the menu.lst entries for the
appropriate BE.
When first setting things up, I attempted to put the proto image onto a
zvol. Part way through this, the machine wedged (and I discovered I had
no working console to it, so I have no more information on that...).
Second time through (after fixing the above, and re-bfu''ing). I
attempted the same, and it worked. I created a file-backed domU, booted
it, and up it came. Nice.
I had to boot the 64bit domU when running on a 64bit dom0. Is this The
Way Things Are, or a current limitation? (errors received from xm
weren''t constant, Bad File Handle was the most common, but there were
others I sadly don''t have a record of).
At that point, I tried creating a second domU, this time on a zvol.
Which wedged the machine, this time I had a console. The machine was
wedged solidly, I couldn''t enter kmdb, 3 C-a''s didn''t
do anything. I
was left having to cycle the machine. I do intend to keep trying to
reproduce this until I can gather some useful information, but as the
machine isn''t local, that may not be a quick affair.
Hardware is dual opteron 240''s, 1G ram.
For the first run, where creating the proto on a zvol went bad, the
machine had 512M swap on disk, 512M on a zvol. Second time through I
increased that to 2G swap on disk, 512M on zvol.
I''ll file CR''s for the two issues above later today, and ask
for them to
be triaged into kernel/xen.
-- Rich.