[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] networking



> I have domains > 0 booting under the latest release. Here's some things I've 
> noticed:
> 
> . even though I have /etc/network/interfaces configured, eth0
> just starts as dhcp, even though there appears to be no dhcp
> daemon running. Maybe this is because it is running in runlevel
> 4?

I think this must be something specific about your setup -- we're
not explicitly running dhcp.

In my Xen installations, I just use the distribution's normal run
level rather than doing anything special. Since we now attach
console devices to tty1 etc rather than xen/cons the
distribution's normal startup scripts just work (at least for RH9
-- I'd be interested to hear if there are problems with other distros).

> . Domain0 has (or will have) eth0, eth1, and eth2. I want
> vif1.0 bridged to eth2, and vif1.1 bridged to eth1. I don't
> seem to get any say in the bridging done in domain0, xend
> appears to do it all itself.

Yes -- this is the current situation, but you should get the
functionality you need back later in the week: you'll be able to
run an arbitrary script to do the bridge plumbing.

> . I can't get any access to dom1's console. telnet localhost
> 9601 just causes xend to complain.

I can reproduce -- this worked fine until very recently, so we
must have had a bad patch.

> . even though I have specified mac addresses in the config
> file, it just gives a random one with the AA prefix. 

I've never tested this, but it sounds like a straightforward bug.

> I think a locally assigned mac address should look like:
> FE:FD:xx:xx:xx:xx. A suggestion for assigning these dynamically
> could be: FE:FD:rr:rr:vv:ii where rr = a random prefix (or
> maybe in a config file), vv is the vmid, ii is the interface
> index. rr:rr would be necessary in a site with multiple xend
> servers.

I think our original idea was to follow the scheme VMWare uses
for assigning them.

For a number of reasons, I think we're going to have to introduce
the notion of a VM Identifier that is unique across machines in a
cluster. This is important for things like migration.

> . Destroying a domain doesn't appear to remove it from the list
> in "xm list"

This wasn't working for quite a while, because the tools hadn't
caught up with the new Xen interface. Mike fixed it on Friday
afternoon but the changes haven't been pushed out yet.

> . When I restarted a domain with 2 nics (had 1 before), dom0 is
> no longer reachable (i'm sitting at home while the server is at
> work so I can't check why at the moment - the server could have
> been stolen for all I can tell from here :).

I haven't tested multiple vifs. Mike relies on them, so I'm sure
this will be made to work.

> As i've mentioned before, i'd really like to see xend call a
> script when a vif comes up, so it can be bridged, routed,
> whatever, as appropriate. Ditto for when a vif goes down.

Yep, you'll get this soon.

> Hope that the above is useful rather than annoying!

Very useful, thanks. I'm hopeful that by the end of next week
we'll be able to declare a 2.0-alpha and hopefully get some wider
testing and comment.

Thanks,
Ian
 -=- MIME -=- 
--_9D85A662-503B-4847-BDF8-39A6C39AD2E2_
Content-Type: text/plain;
        charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

I have domains > 0 booting under the latest release. Here's some things I'v=
e noticed:

. even though I have /etc/network/interfaces configured, eth0 just starts a=
s dhcp, even though there appears to be no dhcp daemon running. Maybe this =
is because it is running in runlevel 4?
. Domain0 has (or will have) eth0, eth1, and eth2. I want vif1.0 bridged to=
 eth2, and vif1.1 bridged to eth1. I don't seem to get any say in the bridg=
ing done in domain0, xend appears to do it all itself.
. I can't get any access to dom1's console. telnet localhost 9601 just caus=
es xend to complain.
. even though I have specified mac addresses in the config file, it just gi=
ves a random one with the AA prefix. I think a locally assigned mac address=
 should look like: FE:FD:xx:xx:xx:xx. A suggestion for assigning these dyna=
mically could be: FE:FD:rr:rr:vv:ii where rr =3D a random prefix (or maybe =
in a config file), vv is the vmid, ii is the interface index. rr:rr would b=
e necessary in a site with multiple xend servers.
. Destroying a domain doesn't appear to remove it from the list in "xm list=
"
. When I restarted a domain with 2 nics (had 1 before), dom0 is no longer r=
eachable (i'm sitting at home while the server is at work so I can't check =
why at the moment - the server could have been stolen for all I can tell fr=
om here :).

As i've mentioned before, i'd really like to see xend call a script when a =
vif comes up, so it can be bridged, routed, whatever, as appropriate. Ditto=
 for when a vif goes down.

Hope that the above is useful rather than annoying!

James

--_9D85A662-503B-4847-BDF8-39A6C39AD2E2_
Content-Type: text/html;
        charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

<HTML dir=3Dltr><HEAD></HEAD>
<BODY>
<DIV><FONT face=3DArial color=3D#000000 size=3D2>I have&nbsp;domains &gt; 0=
 booting&nbsp;under the latest release. Here's some things I've noticed:</F=
ONT></DIV>
<DIV>&nbsp;</DIV>
<DIV><FONT face=3DArial size=3D2>. even though I have /etc/network/interfac=
es configured, eth0 just starts as dhcp, even though there appears to be no=
 dhcp daemon running. Maybe this is because it is running in runlevel 4?</F=
ONT></DIV>
<DIV><FONT face=3DArial size=3D2>. Domain0 has (or will have) eth0, eth1, a=
nd eth2. I want vif1.0 bridged to eth2, and vif1.1 bridged to eth1. I don't=
 seem to get any say in the bridging done in domain0, xend appears to do it=
 all itself.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2>. I can't get any access to dom1's console=
. telnet localhost 9601 just causes xend to complain.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2>. even though I have specified mac address=
es in the config file, it just gives a random one with the AA prefix. I thi=
nk a locally assigned mac address should look like: FE:FD:xx:xx:xx:xx. A su=
ggestion for assigning these dynamically could be: FE:FD:rr:rr:vv:ii where =
rr =3D a random prefix (or maybe in a config file), vv is the vmid, ii is t=
he interface index. rr:rr would be necessary in a site with multiple xend s=
ervers.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2>. Destroying a domain doesn't appear to re=
move it from the list in "xm list"</FONT></DIV>
<DIV><FONT face=3DArial size=3D2>. When I restarted a domain with 2 nics (h=
ad 1 before), dom0 is no longer reachable (i'm sitting at home while the se=
rver is at work so I can't check why at the moment - the server could have =
been stolen for&nbsp;all I can tell from here :).</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial size=3D2>As i've mentioned before, i'd really like =
to see xend call a script when a vif comes up, so it can be bridged, routed=
, whatever, as appropriate. Ditto for when a vif goes down.</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial size=3D2>Hope that the above is useful rather than =
annoying!</FONT></DIV>
<DIV><FONT face=3DArial size=3D2></FONT>&nbsp;</DIV>
<DIV><FONT face=3DArial size=3D2>James</FONT></DIV></BODY></HTML>

--_9D85A662-503B-4847-BDF8-39A6C39AD2E2_--


-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/xen-devel



-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.