From deshantm at gmail.com  Thu Mar  4 11:18:32 2010
From: deshantm at gmail.com (Todd Deshane)
Date: Thu, 4 Mar 2010 11:18:32 -0500
Subject: [ovs-discuss] Database configuration question (Was Re: Open vSwitch
	0.99.2 Available)
Message-ID: <1e16a9ed1003040818k4d231130vb4f5a14fee727b3f@mail.gmail.com>

On Fri, Feb 19, 2010 at 11:56 AM, Justin Pettit <jpettit at nicira.com> wrote:
> Ben recently updated them in the "next" branch, and we're using them internally. ?The "next" branch is what all future "stable" releases will be based on. ?It uses a configuration database instead of a configuration file, so it's not really backwards-compatible from an end-user perspective. ?What time frame are you working on? ?The "next" branch isn't stable enough for production use yet, but basing a release on master may be confusing to users in the future.

Will the database get rid of the need to kill -HUP the switch daemon?

I'm working on a network wrapper interface for another project and I
just wanted to get a basic sense for what I functionality it will need
and how it will make use of the vswitches. Are there other changes
related to basic usage that come along with this new configuration
database management?

Thanks,
Todd

-- 
Todd Deshane
http://todddeshane.net
http://runningxen.com


From blp at nicira.com  Thu Mar  4 12:17:40 2010
From: blp at nicira.com (Ben Pfaff)
Date: Thu, 4 Mar 2010 09:17:40 -0800
Subject: [ovs-discuss] Database configuration question (Was Re: Open
 vSwitch 0.99.2 Available)
In-Reply-To: <1e16a9ed1003040818k4d231130vb4f5a14fee727b3f@mail.gmail.com>
References: <1e16a9ed1003040818k4d231130vb4f5a14fee727b3f@mail.gmail.com>
Message-ID: <20100304171740.GA5448@nicira.com>

On Thu, Mar 04, 2010 at 11:18:32AM -0500, Todd Deshane wrote:
> On Fri, Feb 19, 2010 at 11:56 AM, Justin Pettit <jpettit at nicira.com> wrote:
> > Ben recently updated them in the "next" branch, and we're using them internally. ?The "next" branch is what all future "stable" releases will be based on. ?It uses a configuration database instead of a configuration file, so it's not really backwards-compatible from an end-user perspective. ?What time frame are you working on? ?The "next" branch isn't stable enough for production use yet, but basing a release on master may be confusing to users in the future.

There's a lot of text on the line above.  I think some of it might be
as-yet-unanswered questions, but it's too jumbled to tell.

> Will the database get rid of the need to kill -HUP the switch daemon?

Yes.

> I'm working on a network wrapper interface for another project and I
> just wanted to get a basic sense for what I functionality it will need
> and how it will make use of the vswitches. Are there other changes
> related to basic usage that come along with this new configuration
> database management?

You will need to start the database server daemon along with
ovs-vswitchd, and use the ovs-vsctl utility (which has been greatly
extended) for most configuration tasks.


From deshantm at gmail.com  Mon Mar 22 20:10:44 2010
From: deshantm at gmail.com (Todd Deshane)
Date: Mon, 22 Mar 2010 20:10:44 -0400
Subject: [ovs-discuss] Open vswitch in a driver domain with PCI Passthrough
Message-ID: <1e16a9ed1003221710n173945dby931c786d7cf8669c@mail.gmail.com>

Hi,

Would it be possible to use open vswitch in a driver domain
(Xen or KVM PCI Passthrough) in order to do all the networking outside of dom0?

I couldn't think of a fundamentally limitation that would disallow it...

Also, similarly, is there, or could there be in the future, a way for
an open vswitch to
be used over the network? Meaning that a computer or VM connects over a special
channel to get its networking. Kind of link how an open flow
controller can be used
on the network; I wonder if it is, or ever would be, possible to use
an open vswitch
over the network?

Thanks,
Todd

-- 
Todd Deshane
http://todddeshane.net
http://runningxen.com


From blp at nicira.com  Tue Mar 23 12:29:13 2010
From: blp at nicira.com (Ben Pfaff)
Date: Tue, 23 Mar 2010 09:29:13 -0700
Subject: [ovs-discuss] config/build errors with 2.6.33
In-Reply-To: <9230b3631003230150x18afe2c9q11a8ba97222df55d@mail.gmail.com>
References: <9230b3631003230150x18afe2c9q11a8ba97222df55d@mail.gmail.com>
Message-ID: <20100323162913.GA17971@nicira.com>

On Tue, Mar 23, 2010 at 01:50:52AM -0700, Brandon Heller wrote:
> I had some trouble configuring and building OVS with a custom 2.6.33 kernel.
>  Ugly workaround patch below, which should really use a greater-than
> comparison in Makefile.main.in.  Somewhere after 2.6.29 autconf.h moved
> from include/linux to /include/generated.
> 
> The patch descends from commit 973eab32 (yesterday's OVS master branch).

Thank you for the report.

I think it's easier to just look for the file in both places.  This
works for me on 2.6.32 and 2.6.33 (and presumably other versions):

--8<--------------------------cut here-------------------------->8--

>From 0ccd7844bbe9f6014d162c573e796c8c6f3f707b Mon Sep 17 00:00:00 2001
From: Ben Pfaff <blp at nicira.com>
Date: Tue, 23 Mar 2010 09:27:08 -0700
Subject: [PATCH] datapath: Support 2.6.33 kernel layout in build system.

The 2.6.33 kernel moves generated header files to include/generated, so
we need to look for autoconf.h there.

Reported-by: Brandon Heller <brandonh at stanford.edu>
---
 acinclude.m4                        |    5 +++--
 datapath/linux-2.6/Makefile.main.in |    7 +++++--
 2 files changed, 8 insertions(+), 4 deletions(-)

diff --git a/acinclude.m4 b/acinclude.m4
index e38676f..abbc57e 100644
--- a/acinclude.m4
+++ b/acinclude.m4
@@ -70,8 +70,9 @@ AC_DEFUN([OVS_CHECK_LINUX26], [
          AC_ERROR([Linux kernel in build tree $KBUILD26 (source tree $KSRC26) is not version 2.6])
        fi
     fi
-    if ! test -e "$KBUILD26"/include/linux/version.h || \
-       ! test -e "$KBUILD26"/include/linux/autoconf.h; then
+    if test ! -e "$KBUILD26"/include/linux/version.h || \
+       (test ! -e "$KBUILD26"/include/linux/autoconf.h && \
+        test ! -e "$KBUILD26"/include/generated/autoconf.h); then
 	AC_MSG_ERROR([Linux kernel source in $KBUILD26 is not configured])
     fi
     OVS_CHECK_LINUX26_COMPAT
diff --git a/datapath/linux-2.6/Makefile.main.in b/datapath/linux-2.6/Makefile.main.in
index fd231e6..cb2a042 100644
--- a/datapath/linux-2.6/Makefile.main.in
+++ b/datapath/linux-2.6/Makefile.main.in
@@ -42,9 +42,12 @@ ifeq (,$(wildcard $(VERSION_FILE)))
   $(error Linux kernel source not configured - missing version.h)
 endif
 
-CONFIG_FILE := $(KSRC)/include/linux/autoconf.h
+CONFIG_FILE := $(KSRC)/include/generated/autoconf.h
 ifeq (,$(wildcard $(CONFIG_FILE)))
-  $(error Linux kernel source not configured - missing autoconf.h)
+  CONFIG_FILE := $(KSRC)/include/linux/autoconf.h
+  ifeq (,$(wildcard $(CONFIG_FILE)))
+    $(error Linux kernel source not configured - missing autoconf.h)
+  endif
 endif
 
 default:
-- 
1.6.6.1



From blp at nicira.com  Tue Mar 23 12:40:05 2010
From: blp at nicira.com (Ben Pfaff)
Date: Tue, 23 Mar 2010 09:40:05 -0700
Subject: [ovs-discuss] Open vswitch in a driver domain with PCI
 Passthrough
In-Reply-To: <1e16a9ed1003221710n173945dby931c786d7cf8669c@mail.gmail.com>
References: <1e16a9ed1003221710n173945dby931c786d7cf8669c@mail.gmail.com>
Message-ID: <20100323164005.GD17971@nicira.com>

On Mon, Mar 22, 2010 at 08:10:44PM -0400, Todd Deshane wrote:
> Would it be possible to use open vswitch in a driver domain (Xen or
> KVM PCI Passthrough) in order to do all the networking outside of
> dom0?
> 
> I couldn't think of a fundamentally limitation that would disallow it...

I don't know too much about driver domains.   Theory, yes; practice,
no.

If you can run the Linux bridge in a driver domain, then you should be
able to run Open vSwitch in a driver domain also.

> Also, similarly, is there, or could there be in the future, a way for
> an open vswitch to be used over the network? Meaning that a computer
> or VM connects over a special channel to get its networking. Kind of
> link how an open flow controller can be used on the network; I wonder
> if it is, or ever would be, possible to use an open vswitch over the
> network?

If you are looking for OpenFlow, then Open vSwitch supports OpenFlow.
Beyond that, if you want to configure Open vSwitch over the network,
Open vSwitch exports its configuration database over a custom protocol
called OVSDB that you can access over the network.  Of course, you can
also just ssh in.


From blp at nicira.com  Tue Mar 23 15:20:33 2010
From: blp at nicira.com (Ben Pfaff)
Date: Tue, 23 Mar 2010 12:20:33 -0700
Subject: [ovs-discuss] config/build errors with 2.6.33
In-Reply-To: <9230b3631003231207k54658419wbcdb5d9f7dc24d8b@mail.gmail.com>
References: <9230b3631003230150x18afe2c9q11a8ba97222df55d@mail.gmail.com>
	<20100323162913.GA17971@nicira.com>
	<9230b3631003231207k54658419wbcdb5d9f7dc24d8b@mail.gmail.com>
Message-ID: <20100323192033.GK17971@nicira.com>

Thanks for testing it.

I pushed this to OVS "master".

On Tue, Mar 23, 2010 at 12:07:04PM -0700, Brandon Heller wrote:
> Hi Ben,
> 
> This works on my system and is much cleaner.  Thanks!
> 
> -Brandon
> 
> On Tue, Mar 23, 2010 at 9:29 AM, Ben Pfaff <blp at nicira.com> wrote:
> 
> > On Tue, Mar 23, 2010 at 01:50:52AM -0700, Brandon Heller wrote:
> > > I had some trouble configuring and building OVS with a custom 2.6.33
> > kernel.
> > >  Ugly workaround patch below, which should really use a greater-than
> > > comparison in Makefile.main.in.  Somewhere after 2.6.29 autconf.h moved
> > > from include/linux to /include/generated.
> > >
> > > The patch descends from commit 973eab32 (yesterday's OVS master branch).
> >
> > Thank you for the report.
> >
> > I think it's easier to just look for the file in both places.  This
> > works for me on 2.6.32 and 2.6.33 (and presumably other versions):
> >
> > --8<--------------------------cut here-------------------------->8--
> >
> > >From 0ccd7844bbe9f6014d162c573e796c8c6f3f707b Mon Sep 17 00:00:00 2001
> > From: Ben Pfaff <blp at nicira.com>
> > Date: Tue, 23 Mar 2010 09:27:08 -0700
> > Subject: [PATCH] datapath: Support 2.6.33 kernel layout in build system.
> >
> > The 2.6.33 kernel moves generated header files to include/generated, so
> > we need to look for autoconf.h there.
> >
> > Reported-by: Brandon Heller <brandonh at stanford.edu>
> > ---
> >  acinclude.m4                        |    5 +++--
> >  datapath/linux-2.6/Makefile.main.in |    7 +++++--
> >  2 files changed, 8 insertions(+), 4 deletions(-)
> >
> > diff --git a/acinclude.m4 b/acinclude.m4
> > index e38676f..abbc57e 100644
> > --- a/acinclude.m4
> > +++ b/acinclude.m4
> > @@ -70,8 +70,9 @@ AC_DEFUN([OVS_CHECK_LINUX26], [
> >          AC_ERROR([Linux kernel in build tree $KBUILD26 (source tree
> > $KSRC26) is not version 2.6])
> >         fi
> >     fi
> > -    if ! test -e "$KBUILD26"/include/linux/version.h || \
> > -       ! test -e "$KBUILD26"/include/linux/autoconf.h; then
> > +    if test ! -e "$KBUILD26"/include/linux/version.h || \
> > +       (test ! -e "$KBUILD26"/include/linux/autoconf.h && \
> > +        test ! -e "$KBUILD26"/include/generated/autoconf.h); then
> >        AC_MSG_ERROR([Linux kernel source in $KBUILD26 is not configured])
> >     fi
> >      OVS_CHECK_LINUX26_COMPAT
> > diff --git a/datapath/linux-2.6/Makefile.main.in b/datapath/linux-2.6/
> > Makefile.main.in
> > index fd231e6..cb2a042 100644
> > --- a/datapath/linux-2.6/Makefile.main.in
> > +++ b/datapath/linux-2.6/Makefile.main.in
> > @@ -42,9 +42,12 @@ ifeq (,$(wildcard $(VERSION_FILE)))
> >    $(error Linux kernel source not configured - missing version.h)
> >  endif
> >
> > -CONFIG_FILE := $(KSRC)/include/linux/autoconf.h
> > +CONFIG_FILE := $(KSRC)/include/generated/autoconf.h
> >  ifeq (,$(wildcard $(CONFIG_FILE)))
> > -  $(error Linux kernel source not configured - missing autoconf.h)
> > +  CONFIG_FILE := $(KSRC)/include/linux/autoconf.h
> > +  ifeq (,$(wildcard $(CONFIG_FILE)))
> > +    $(error Linux kernel source not configured - missing autoconf.h)
> > +  endif
> >  endif
> >
> >  default:
> > --
> > 1.6.6.1
> >
> >
> > _______________________________________________
> > discuss mailing list
> > discuss at openvswitch.org
> > http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org
> >


From blp at nicira.com  Tue Mar 23 17:58:59 2010
From: blp at nicira.com (Ben Pfaff)
Date: Tue, 23 Mar 2010 14:58:59 -0700
Subject: [ovs-discuss] Fwd: OVS 0.90.7 build fails on kernel
 2.6.31-14-generic
In-Reply-To: <f744c9851003231405j589c2107n63b83b2867f00ca8@mail.gmail.com>
References: <f744c9851003231247w10f33085nfac95e35e2d88662@mail.gmail.com>
	<25cfcaf91003231319x66a48b8cl667f9fc871b2793e@mail.gmail.com>
	<f744c9851003231356o3048596dmb070a2d0bae05677@mail.gmail.com>
	<f744c9851003231405j589c2107n63b83b2867f00ca8@mail.gmail.com>
Message-ID: <20100323215859.GA30017@nicira.com>

On Tue, Mar 23, 2010 at 02:05:56PM -0700, Prateek Jain wrote:
> Thanks for your reply. I tried downloading and building 0.99.2 in the
> mean time.  I tried building for kernel modules, the build was
> successful, however the loading of kernel module fails with the
> following error
> 
> insmod: error inserting './datapath/linux-2.6/openvswitch_mod.ko': -1
> Invalid module format.

Probably the compiler you used to build the module is different from the
compiler used to build the kernel.  Compare the "vermagic" output by
/sbin/modinfo for openvswitch_mod.ko against that output for a module
installed in /lib/modules/`uname r`.


From blp at nicira.com  Tue Mar 23 19:48:36 2010
From: blp at nicira.com (Ben Pfaff)
Date: Tue, 23 Mar 2010 16:48:36 -0700
Subject: [ovs-discuss] Fwd: OVS 0.90.7 build fails on kernel
 2.6.31-14-generic
In-Reply-To: <f744c9851003231627s5e19f942o5abb592f8853a0a@mail.gmail.com>
References: <f744c9851003231247w10f33085nfac95e35e2d88662@mail.gmail.com>
	<25cfcaf91003231319x66a48b8cl667f9fc871b2793e@mail.gmail.com>
	<f744c9851003231356o3048596dmb070a2d0bae05677@mail.gmail.com>
	<f744c9851003231405j589c2107n63b83b2867f00ca8@mail.gmail.com>
	<20100323215859.GA30017@nicira.com>
	<f744c9851003231627s5e19f942o5abb592f8853a0a@mail.gmail.com>
Message-ID: <20100323234836.GB30017@nicira.com>

On Tue, Mar 23, 2010 at 04:27:38PM -0700, Prateek Jain wrote:
> Thanks, see the attached image, the vermagic output for an already loaded
> module is same as the kernel module complied by me.
> 
> What else could be causing it ?

I can see that you have the "ofdatapath" module from OpenFlow also
loaded.  I don't think that Open vSwitch will work at the same thing as
OpenFlow.  I don't think it would cause this problem, though.

You may be able to find out more information from "dmesg".


From jklee at hp.com  Wed Mar 24 17:27:46 2010
From: jklee at hp.com (Lee, Jeongkeun)
Date: Wed, 24 Mar 2010 21:27:46 +0000
Subject: [ovs-discuss] brcompatd: "could not find valid configuration to
	update" error
Message-ID: <A646510FAEBF484C81F1656322F311DC77C56F18AA@GVW1098EXB.americas.hpqcorp.net>

Hi,

Today, I downloaded the latest snapshot from git and followed the instructions of INSTALL.bridge and INSTALL.Linux to use ovs as a replacement of Linux bridge.
OVS version is 0.99.2 and kernel is 2.6.26-2-amd64 on Debian lenny 5.03.

Below are the commands that I ran after 'make install' completed. ovsdb-server, ovs-vswitchd, and ovs-brcompatd were called in separate shells to see the messages.

% insmod datapath/linux-2.6/openvswitch_mod.ko
% insmod datapath/linux-2.6/brcompat_mod.ko
% ovsdb-tool create /etc/ovs-vswitchd.conf.db vswitchd/vswitch.ovsschema
% ovsdb-server /etc/ovs-vswitchd.conf.db --remote=punix:/var/run/ovsdb-server
% ovs-vswitchd -vANY:console:WARN unix:/var/run/ovsdb-server
% ovs-brcompatd -vANY:console:WARN unix:/var/run/ovsdb-server

ovsdb-server shows messages saying ovs-vswitchd and ovs-brcompatd are connected 
and ovs-brcompatd complains 
	database does not contain any Open vSwitch configuration
when it gets started.

Then, 
	% brctl addbr br0
results in ovs-brcompatd dumping messages like
	Mar 24 12:16:48|00031|brcompatd|WARN|could not find valid configuration to update
	Mar 24 12:17:21|00032|brcompatd|WARN|Dropped 32 log messages in last 60 seconds due to excessive rate
	Mar 24 12:17:21|00033|brcompatd|WARN|unix:/var/run/ovsdb-server: database does not contain any Open vSwitch configuration
and the brctl stops with 
	add bridge failed: Connection timed out

It seems like I should fill up vswitchd/vswitch.ovsschema with right configurations before calling 'ovsdb-tool create', should I? 

Btw, ovs-vsctl connects to /usr/local/var/run/ovsdb-server by default instead of /var/run/ovsdb-server. Either ovs-vsctl or the INSTALL instruction needs to be revised to remove the discrepancy. 

Thanks,
-- JK




From jklee at hp.com  Wed Mar 24 17:33:49 2010
From: jklee at hp.com (Lee, Jeongkeun)
Date: Wed, 24 Mar 2010 21:33:49 +0000
Subject: [ovs-discuss] brcompatd: "could not find valid configuration
	to	update" error
In-Reply-To: <A646510FAEBF484C81F1656322F311DC77C56F18AA@GVW1098EXB.americas.hpqcorp.net>
References: <A646510FAEBF484C81F1656322F311DC77C56F18AA@GVW1098EXB.americas.hpqcorp.net>
Message-ID: <A646510FAEBF484C81F1656322F311DC77C56F18B6@GVW1098EXB.americas.hpqcorp.net>

calling 'ovs-vsctl init' solved the issue.

thanks,
-- JK

> -----Original Message-----
> From: discuss-bounces at openvswitch.org [mailto:discuss-
> bounces at openvswitch.org] On Behalf Of Lee, Jeongkeun
> Sent: Wednesday, March 24, 2010 2:28 PM
> To: discuss at openvswitch.org
> Subject: [ovs-discuss] brcompatd: "could not find valid configuration to
> update" error
> 
> Hi,
> 
> Today, I downloaded the latest snapshot from git and followed the
> instructions of INSTALL.bridge and INSTALL.Linux to use ovs as a
> replacement of Linux bridge.
> OVS version is 0.99.2 and kernel is 2.6.26-2-amd64 on Debian lenny 5.03.
> 
> Below are the commands that I ran after 'make install' completed. ovsdb-
> server, ovs-vswitchd, and ovs-brcompatd were called in separate shells to
> see the messages.
> 
> % insmod datapath/linux-2.6/openvswitch_mod.ko
> % insmod datapath/linux-2.6/brcompat_mod.ko
> % ovsdb-tool create /etc/ovs-vswitchd.conf.db vswitchd/vswitch.ovsschema
> % ovsdb-server /etc/ovs-vswitchd.conf.db --remote=punix:/var/run/ovsdb-
> server
> % ovs-vswitchd -vANY:console:WARN unix:/var/run/ovsdb-server
> % ovs-brcompatd -vANY:console:WARN unix:/var/run/ovsdb-server
> 
> ovsdb-server shows messages saying ovs-vswitchd and ovs-brcompatd are
> connected
> and ovs-brcompatd complains
> 	database does not contain any Open vSwitch configuration
> when it gets started.
> 
> Then,
> 	% brctl addbr br0
> results in ovs-brcompatd dumping messages like
> 	Mar 24 12:16:48|00031|brcompatd|WARN|could not find valid
> configuration to update
> 	Mar 24 12:17:21|00032|brcompatd|WARN|Dropped 32 log messages in
> last 60 seconds due to excessive rate
> 	Mar 24 12:17:21|00033|brcompatd|WARN|unix:/var/run/ovsdb-server:
> database does not contain any Open vSwitch configuration
> and the brctl stops with
> 	add bridge failed: Connection timed out
> 
> It seems like I should fill up vswitchd/vswitch.ovsschema with right
> configurations before calling 'ovsdb-tool create', should I?
> 
> Btw, ovs-vsctl connects to /usr/local/var/run/ovsdb-server by default
> instead of /var/run/ovsdb-server. Either ovs-vsctl or the INSTALL
> instruction needs to be revised to remove the discrepancy.
> 
> Thanks,
> -- JK
> 
> 
> 
> _______________________________________________
> discuss mailing list
> discuss at openvswitch.org
> http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org


From blp at nicira.com  Wed Mar 24 17:44:13 2010
From: blp at nicira.com (Ben Pfaff)
Date: Wed, 24 Mar 2010 14:44:13 -0700
Subject: [ovs-discuss] brcompatd: "could not find valid configuration to
 update" error
In-Reply-To: <A646510FAEBF484C81F1656322F311DC77C56F18AA@GVW1098EXB.americas.hpqcorp.net>
References: <A646510FAEBF484C81F1656322F311DC77C56F18AA@GVW1098EXB.americas.hpqcorp.net>
Message-ID: <20100324214413.GF7368@nicira.com>

On Wed, Mar 24, 2010 at 09:27:46PM +0000, Lee, Jeongkeun wrote:
> Today, I downloaded the latest snapshot from git and followed the
> instructions of INSTALL.bridge and INSTALL.Linux to use ovs as a
> replacement of Linux bridge.  OVS version is 0.99.2 and kernel is
> 2.6.26-2-amd64 on Debian lenny 5.03.
> 
> Below are the commands that I ran after 'make install'
> completed. ovsdb-server, ovs-vswitchd, and ovs-brcompatd were called
> in separate shells to see the messages.
> 
> % insmod datapath/linux-2.6/openvswitch_mod.ko
> % insmod datapath/linux-2.6/brcompat_mod.ko
> % ovsdb-tool create /etc/ovs-vswitchd.conf.db vswitchd/vswitch.ovsschema
> % ovsdb-server /etc/ovs-vswitchd.conf.db --remote=punix:/var/run/ovsdb-server
> % ovs-vswitchd -vANY:console:WARN unix:/var/run/ovsdb-server
> % ovs-brcompatd -vANY:console:WARN unix:/var/run/ovsdb-server
> 
> ovsdb-server shows messages saying ovs-vswitchd and ovs-brcompatd are connected 
> and ovs-brcompatd complains 
> 	database does not contain any Open vSwitch configuration
> when it gets started.

Thanks for reporting this problem.

You're right, there's a missing step here.  I'm going to add this
paragraph to INSTALL.Linux:

    Then initialize the database with "ovs-vsctl init".  This is only
    necessary the first time after you create the database with
    ovsdb-tool (but running it at any time is harmless):

          % ovs-vsctl init

after starting ovsdb-server and before starting ovs-vswitchd.
 
> Btw, ovs-vsctl connects to /usr/local/var/run/ovsdb-server by default
> instead of /var/run/ovsdb-server. Either ovs-vsctl or the INSTALL
> instruction needs to be revised to remove the discrepancy.

You are right again, of course.  The documentation is inconsistent here.
It assumes / in some places and /usr/local in others.  I will also push
out this fix:

    diff --git a/INSTALL.Linux b/INSTALL.Linux
    index 31bc3e0..263e261 100644
    --- a/INSTALL.Linux
    +++ b/INSTALL.Linux
    @@ -157,7 +157,7 @@ Prerequisites section, follow the procedure below to build.

     7. Initialize the configuration database using ovsdb-tool, e.g.:

    -      % ovsdb-tool create /etc/ovs-vswitchd.conf.db vswitchd/vswitch.ovsschema
    +      % ovsdb-tool create /usr/local/etc/ovs-vswitchd.conf.db vswitchd/vswitch.ovsschema

     Startup
     =======
    @@ -167,7 +167,7 @@ configuration database, ovsdb-server.  Configure it to use the
     database you created during step 7 of installation, above, and to
     listen on a Unix domain socket, e.g.:

    -      % ovsdb-server /etc/ovs-vswitchd.conf.db --remote=punix:/var/run/ovsdb-server
    +      % ovsdb-server /usr/local/etc/ovs-vswitchd.conf.db --remote=punix:/usr/local/var/run/ovsdb-server

     Then initialize the database with "ovs-vsctl init".  This is only
     necessary the first time after you create the database with
    @@ -178,7 +178,7 @@ ovsdb-tool (but running it at any time is harmless):
     Then start the main Open vSwitch daemon, telling it to connect to the
     same Unix domain socket:

    -      % ovs-vswitchd unix:/var/run/ovsdb-server
    +      % ovs-vswitchd unix:/usr/local/var/run/ovsdb-server

     Now you may use ovs-vsctl to set up bridges and other Open vSwitch
     features.  For example, to create a bridge named br0 and add ports

INSTALL.Linux does explain how to change the default:

   By default all files are installed under /usr/local.  If you want
   to install into, e.g., /usr and /var instead of /usr/local and
   /usr/local/var, add options as shown here:

      % ./configure --prefix=/usr --localstatedir=/var


From blp at nicira.com  Wed Mar 24 17:46:36 2010
From: blp at nicira.com (Ben Pfaff)
Date: Wed, 24 Mar 2010 14:46:36 -0700
Subject: [ovs-discuss] brcompatd: "could not find valid configuration to
 update" error
In-Reply-To: <A646510FAEBF484C81F1656322F311DC77C56F18AA@GVW1098EXB.americas.hpqcorp.net>
References: <A646510FAEBF484C81F1656322F311DC77C56F18AA@GVW1098EXB.americas.hpqcorp.net>
Message-ID: <20100324214636.GG7368@nicira.com>

On Wed, Mar 24, 2010 at 09:27:46PM +0000, Lee, Jeongkeun wrote:
> It seems like I should fill up vswitchd/vswitch.ovsschema with right
> configurations before calling 'ovsdb-tool create', should I?

I just noticed this part of your message.

The ovsschema file is fixed for any given version of Open vSwitch.  It
is read-only and used only for creating the database.  It's not used by
ovsdb-server at all but only in the "ovsdb-tool create" step.

It's the database file (which we've mostly been naming
/etc/ovs-vswitchd.conf.db but that's an awful name and we're planning to
change it) that changes.


From jklee at hp.com  Wed Mar 24 22:51:44 2010
From: jklee at hp.com (Lee, Jeongkeun)
Date: Thu, 25 Mar 2010 02:51:44 +0000
Subject: [ovs-discuss] brcompatd: "could not find valid configuration to
 update" error
In-Reply-To: <20100324214413.GF7368@nicira.com>
References: <A646510FAEBF484C81F1656322F311DC77C56F18AA@GVW1098EXB.americas.hpqcorp.net>
	<20100324214413.GF7368@nicira.com>
Message-ID: <A646510FAEBF484C81F1656322F311DC77C56F19DE@GVW1098EXB.americas.hpqcorp.net>

Ben,

Thanks for the quick answers.
I think INSTALL.bridge also needs to be fixed to use /usr/local.

thanks,
JK

> -----Original Message-----
> From: Ben Pfaff [mailto:blp at nicira.com]
> Sent: Wednesday, March 24, 2010 2:44 PM
> To: Lee, Jeongkeun
> Cc: discuss at openvswitch.org
> Subject: Re: [ovs-discuss] brcompatd: "could not find valid configuration
> to update" error
> 
> On Wed, Mar 24, 2010 at 09:27:46PM +0000, Lee, Jeongkeun wrote:
> > Today, I downloaded the latest snapshot from git and followed the
> > instructions of INSTALL.bridge and INSTALL.Linux to use ovs as a
> > replacement of Linux bridge.  OVS version is 0.99.2 and kernel is
> > 2.6.26-2-amd64 on Debian lenny 5.03.
> >
> > Below are the commands that I ran after 'make install'
> > completed. ovsdb-server, ovs-vswitchd, and ovs-brcompatd were called
> > in separate shells to see the messages.
> >
> > % insmod datapath/linux-2.6/openvswitch_mod.ko
> > % insmod datapath/linux-2.6/brcompat_mod.ko
> > % ovsdb-tool create /etc/ovs-vswitchd.conf.db
> vswitchd/vswitch.ovsschema
> > % ovsdb-server /etc/ovs-vswitchd.conf.db --remote=punix:/var/run/ovsdb-
> server
> > % ovs-vswitchd -vANY:console:WARN unix:/var/run/ovsdb-server
> > % ovs-brcompatd -vANY:console:WARN unix:/var/run/ovsdb-server
> >
> > ovsdb-server shows messages saying ovs-vswitchd and ovs-brcompatd are
> connected
> > and ovs-brcompatd complains
> > 	database does not contain any Open vSwitch configuration
> > when it gets started.
> 
> Thanks for reporting this problem.
> 
> You're right, there's a missing step here.  I'm going to add this
> paragraph to INSTALL.Linux:
> 
>     Then initialize the database with "ovs-vsctl init".  This is only
>     necessary the first time after you create the database with
>     ovsdb-tool (but running it at any time is harmless):
> 
>           % ovs-vsctl init
> 
> after starting ovsdb-server and before starting ovs-vswitchd.
> 
> > Btw, ovs-vsctl connects to /usr/local/var/run/ovsdb-server by default
> > instead of /var/run/ovsdb-server. Either ovs-vsctl or the INSTALL
> > instruction needs to be revised to remove the discrepancy.
> 
> You are right again, of course.  The documentation is inconsistent here.
> It assumes / in some places and /usr/local in others.  I will also push
> out this fix:
> 
>     diff --git a/INSTALL.Linux b/INSTALL.Linux
>     index 31bc3e0..263e261 100644
>     --- a/INSTALL.Linux
>     +++ b/INSTALL.Linux
>     @@ -157,7 +157,7 @@ Prerequisites section, follow the procedure below
> to build.
> 
>      7. Initialize the configuration database using ovsdb-tool, e.g.:
> 
>     -      % ovsdb-tool create /etc/ovs-vswitchd.conf.db
> vswitchd/vswitch.ovsschema
>     +      % ovsdb-tool create /usr/local/etc/ovs-vswitchd.conf.db
> vswitchd/vswitch.ovsschema
> 
>      Startup
>      =======
>     @@ -167,7 +167,7 @@ configuration database, ovsdb-server.  Configure
> it to use the
>      database you created during step 7 of installation, above, and to
>      listen on a Unix domain socket, e.g.:
> 
>     -      % ovsdb-server /etc/ovs-vswitchd.conf.db --
> remote=punix:/var/run/ovsdb-server
>     +      % ovsdb-server /usr/local/etc/ovs-vswitchd.conf.db --
> remote=punix:/usr/local/var/run/ovsdb-server
> 
>      Then initialize the database with "ovs-vsctl init".  This is only
>      necessary the first time after you create the database with
>     @@ -178,7 +178,7 @@ ovsdb-tool (but running it at any time is
> harmless):
>      Then start the main Open vSwitch daemon, telling it to connect to
> the
>      same Unix domain socket:
> 
>     -      % ovs-vswitchd unix:/var/run/ovsdb-server
>     +      % ovs-vswitchd unix:/usr/local/var/run/ovsdb-server
> 
>      Now you may use ovs-vsctl to set up bridges and other Open vSwitch
>      features.  For example, to create a bridge named br0 and add ports
> 
> INSTALL.Linux does explain how to change the default:
> 
>    By default all files are installed under /usr/local.  If you want
>    to install into, e.g., /usr and /var instead of /usr/local and
>    /usr/local/var, add options as shown here:
> 
>       % ./configure --prefix=/usr --localstatedir=/var


From jklee at hp.com  Thu Mar 25 00:36:05 2010
From: jklee at hp.com (Lee, Jeongkeun)
Date: Thu, 25 Mar 2010 04:36:05 +0000
Subject: [ovs-discuss] bridge renaming issue with Xen
In-Reply-To: <20100324214636.GG7368@nicira.com>
References: <A646510FAEBF484C81F1656322F311DC77C56F18AA@GVW1098EXB.americas.hpqcorp.net>
	<20100324214636.GG7368@nicira.com>
Message-ID: <A646510FAEBF484C81F1656322F311DC77C56F19FB@GVW1098EXB.americas.hpqcorp.net>

Hi Ben and Jesse,

I'm trying to use ovs 0.99.2 with Xen on Lenny. 
As you know, Xen calls 'ip link set' commands to rename a bridge but ovs-brcompatd does not proxy 'ip link set name' command.

> From: "Jesse Gross" <jesse at nicira.com>
> 
> Earlier versions of Xen networking (which apparently the Lenny port is 
> based
> on) used some "interesting" tricks with renaming. In this case eth0 is 
> renamed to peth0 and a bridge named tmpbridge is created then renamed 
> to eth0.

With the previous version of ovs that doesn't use OVSDB, I simply changed the bridge name directly on the ovs/brcompatd config file by modifying Xen networking scripts as follows:

    ...
    ip link set ${tdev} name ${bridge}
    # ovs-brcompatd does not proxy 'ip link set name' command
    # need to manually change the bridge name in the ovs config file
    # usually tdev=tmpbridge, bridge=eth0
    sed -i "s/${tdev}/${bridge}/g" /etc/ovs-vswitchd-xen.conf  <- added line
    ...

So, the same renaming should be done on the config DB with ovs 0.99.2 and I tried the following 'ovs-vsctl set' commands:
    
    ip link set ${tdev} name ${bridge}
    # sed -i "s/${tdev}/${bridge}/g" ${cfg}
    # usually tdev=tmpbridge, bridge=eth0
    /usr/local/bin/ovs-vsctl set Bridge ${tdev} name=${bridge}
    /usr/local/bin/ovs-vsctl set Port ${tdev} name=${bridge}
    /usr/local/bin/ovs-vsctl set Interface ${tdev} name=${bridge}

Then, after finishing booting, 'brctl show' and 'ovs-dpctl show' look fine in dom0 but generating any packets from domU crashes the system and dom0 automatically reboots. Syslog caught warning and error messages shown below. It seems like 'ovs-vsctl set' does more than just renaming DB entries. Can you help me to find the right use of this command? maybe the order between Bridge, Port, and Interface renaming matters..?



~~~~~~~~~~~~~~~~~~~ SYSLOG dump ~~~~~~~~~~~~~~~~~~~~~~~~~
Mar 24 20:14:52 majuan ovs-vsctl: 00001|vsctl|INFO|Called as /usr/local/bin/ovs-vsctl set Bridge tmpbridge name=eth0
Mar 24 20:14:52 majuan ovsdb-server: 00009|reconnect|INFO|unix:/tmp/stream-unix.4694.0: connecting...
Mar 24 20:14:52 majuan ovsdb-server: 00010|reconnect|INFO|unix:/tmp/stream-unix.4694.0: connected
Mar 24 20:14:52 majuan ovs-brcompatd: 00003|brcompatd|INFO|kernel reported network device eth0 removed but a device by that name exists (XS Tools 5.0.0?)
Mar 24 20:14:52 majuan ovs-vswitchd: 00010|dpif|WARN|system at dp0: flow_flush failed (No such device)
Mar 24 20:14:52 majuan ovs-vswitchd: 00011|ofproto|INFO|using datapath ID 000000232099dc82
Mar 24 20:14:52 majuan ovs-vswitchd: 00012|bridge|INFO|created bridge eth0 on system at dp0
Mar 24 20:14:52 majuan ovs-vswitchd: 00013|bridge|INFO|created port tmpbridge on bridge eth0
Mar 24 20:14:52 majuan ovs-vswitchd: 00014|netdev|WARN|failed to get flags for network device tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00015|bridge|WARN|could not create iface tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00016|dpif|WARN|system at dp0: failed to add tmpbridge as port: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00017|bridge|ERR|failed to add tmpbridge interface to system at dp0: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00018|netdev|WARN|failed to get flags for network device tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00019|bridge|WARN|could not open netdev on tmpbridge, dropping: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00020|bridge|ERR|tmpbridge port has no interfaces, dropping
Mar 24 20:14:52 majuan ovs-vswitchd: 00021|bridge|WARN|bridge eth0: using default bridge Ethernet address 00:23:20:cf:19:dc
Mar 24 20:14:52 majuan ovs-vswitchd: 00022|ofproto|INFO|datapath ID changed to 0000002320cf19dc
Mar 24 20:14:52 majuan ovs-vswitchd: 00023|netdev|WARN|attempted to create a device that may not be created: eth0
Mar 24 20:14:52 majuan ovs-vswitchd: 00024|ofproto|WARN|ignoring port eth0 (0) because netdev eth0 cannot be opened (No such device)
Mar 24 20:14:52 majuan ovsdb-server: 00011|jsonrpc|INFO|unix:/tmp/stream-unix.4694.0: connection closed
Mar 24 20:14:52 majuan ovsdb-server: 00012|reconnect|INFO|unix:/tmp/stream-unix.4694.0: connection dropped
Mar 24 20:14:52 majuan ovs-vswitchd: 00025|netdev|WARN|attempted to create a device that may not be created: eth0
Mar 24 20:14:52 majuan ovs-vswitchd: 00026|ofproto|WARN|ignoring port eth0 (0) because netdev eth0 cannot be opened (No such device)
Mar 24 20:14:52 majuan ovs-vswitchd: 00027|bridge|INFO|created port tmpbridge on bridge eth0

Mar 24 20:14:52 majuan ovs-vsctl: 00001|vsctl|INFO|Called as /usr/local/bin/ovs-vsctl set Port tmpbridge name=eth0
Mar 24 20:14:52 majuan ovsdb-server: 00013|reconnect|INFO|unix:/tmp/stream-unix.4712.0: connecting...
Mar 24 20:14:52 majuan ovsdb-server: 00014|reconnect|INFO|unix:/tmp/stream-unix.4712.0: connected
Mar 24 20:14:52 majuan ovs-vswitchd: 00028|netdev|WARN|failed to get flags for network device tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00029|bridge|WARN|could not create iface tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00030|dpif|WARN|system at dp0: failed to add tmpbridge as port: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00031|bridge|ERR|failed to add tmpbridge interface to system at dp0: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00032|netdev|WARN|failed to get flags for network device tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00033|bridge|WARN|could not open netdev on tmpbridge, dropping: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00034|bridge|ERR|tmpbridge port has no interfaces, dropping
Mar 24 20:14:52 majuan ovs-vswitchd: 00035|bridge|WARN|bridge eth0: using default bridge Ethernet address 00:23:20:cf:19:dc
Mar 24 20:14:52 majuan ovs-vswitchd: 00036|netdev|WARN|attempted to create a device that may not be created: eth0
Mar 24 20:14:52 majuan ovs-vswitchd: 00037|ofproto|WARN|ignoring port eth0 (0) because netdev eth0 cannot be opened (No such device)
Mar 24 20:14:52 majuan ovs-vswitchd: 00038|bridge|INFO|created port eth0 on bridge eth0
Mar 24 20:14:52 majuan ovs-vswitchd: 00039|netdev|WARN|failed to get flags for network device tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00040|bridge|WARN|could not create iface tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00041|dpif|WARN|system at dp0: failed to add tmpbridge as port: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00042|bridge|ERR|failed to add tmpbridge interface to system at dp0: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00043|netdev|WARN|failed to get flags for network device tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00044|bridge|WARN|could not open netdev on tmpbridge, dropping: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00045|bridge|ERR|eth0 port has no interfaces, dropping
Mar 24 20:14:52 majuan ovs-vswitchd: 00046|bridge|WARN|bridge eth0: using default bridge Ethernet address 00:23:20:cf:19:dc
Mar 24 20:14:52 majuan ovs-vswitchd: 00047|netdev|WARN|attempted to create a device that may not be created: eth0
Mar 24 20:14:52 majuan ovs-vswitchd: 00048|ofproto|WARN|ignoring port eth0 (0) because netdev eth0 cannot be opened (No such device)
Mar 24 20:14:52 majuan ovsdb-server: 00015|jsonrpc|INFO|unix:/tmp/stream-unix.4712.0: connection closed
Mar 24 20:14:52 majuan ovsdb-server: 00016|reconnect|INFO|unix:/tmp/stream-unix.4712.0: connection dropped
Mar 24 20:14:52 majuan ovs-vswitchd: 00049|netdev|WARN|attempted to create a device that may not be created: eth0
Mar 24 20:14:52 majuan ovs-vswitchd: 00050|ofproto|WARN|ignoring port eth0 (0) because netdev eth0 cannot be opened (No such device)
Mar 24 20:14:52 majuan ovs-vswitchd: 00051|bridge|INFO|created port eth0 on bridge eth0

Mar 24 20:14:52 majuan ovs-vsctl: 00001|vsctl|INFO|Called as /usr/local/bin/ovs-vsctl set Interface tmpbridge name=eth0
Mar 24 20:14:52 majuan ovsdb-server: 00017|reconnect|INFO|unix:/tmp/stream-unix.4721.0: connecting...
Mar 24 20:14:52 majuan ovsdb-server: 00018|reconnect|INFO|unix:/tmp/stream-unix.4721.0: connected
Mar 24 20:14:52 majuan ovs-vswitchd: 00052|netdev|WARN|failed to get flags for network device tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00053|bridge|WARN|could not create iface tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00054|dpif|WARN|system at dp0: failed to add tmpbridge as port: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00055|bridge|ERR|failed to add tmpbridge interface to system at dp0: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00056|netdev|WARN|failed to get flags for network device tmpbridge: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00057|bridge|WARN|could not open netdev on tmpbridge, dropping: No such device
Mar 24 20:14:52 majuan ovs-vswitchd: 00058|bridge|ERR|eth0 port has no interfaces, dropping
Mar 24 20:14:52 majuan ovs-vswitchd: 00059|bridge|WARN|bridge eth0: using default bridge Ethernet address 00:23:20:cf:19:dc
Mar 24 20:14:52 majuan ovs-vswitchd: 00060|netdev|WARN|attempted to create a device that may not be created: eth0
Mar 24 20:14:52 majuan ovs-vswitchd: 00061|bridge|INFO|created port eth0 on bridge eth0
Mar 24 20:14:52 majuan ovs-vswitchd: 00062|bridge|WARN|bridge eth0: using default bridge Ethernet address 00:23:20:cf:19:dc
Mar 24 20:14:52 majuan ovs-vswitchd: 00063|bridge|WARN|bridge eth0: using default bridge Ethernet address 00:23:20:cf:19:dc
Mar 24 20:14:52 majuan ovsdb-server: 00019|jsonrpc|INFO|unix:/tmp/stream-unix.4721.0: connection closed
Mar 24 20:14:52 majuan ovsdb-server: 00020|reconnect|INFO|unix:/tmp/stream-unix.4721.0: connection dropped
Mar 24 20:14:55 majuan ovs-vswitchd: 00064|bridge|INFO|created port peth0 on bridge eth0
Mar 24 20:14:55 majuan ovs-vswitchd: 00065|ofproto|INFO|datapath ID changed to 000018a905162754
Mar 24 20:14:55 majuan ovs-brcompatd: 00004|brcompatd|INFO|add-if eth0 peth0: Success






From jklee at hp.com  Thu Mar 25 04:10:54 2010
From: jklee at hp.com (Lee, Jeongkeun)
Date: Thu, 25 Mar 2010 08:10:54 +0000
Subject: [ovs-discuss] GRE tunnel with 0.99.2 snapshot
Message-ID: <A646510FAEBF484C81F1656322F311DC77C56F1A3C@GVW1098EXB.americas.hpqcorp.net>

Hi,

Does the current snapshot support GRE tunnel?
Based on the mailing list logs, gre support is under development and will be a part of the next release.

I just tried to attach a gre interface (created by linux 'ip') by ovs-vsctl add-port and ovs-dpctl add-if but none of them worked. I wonder there is any APIs used for testing.

I found a mention about gre interface type in the ovs-vswichtd.conf.db man page, though..

thanks,
-- JK

Jeongkeun "JK" Lee
Hewlett-Packard Labs
Palo Alto, CA




From darkbls at yahoo.com  Thu Mar 25 04:28:34 2010
From: darkbls at yahoo.com (DarkBls)
Date: Thu, 25 Mar 2010 08:28:34 +0000 (GMT)
Subject: [ovs-discuss] Re :  GRE tunnel with 0.99.2 snapshot
In-Reply-To: <A646510FAEBF484C81F1656322F311DC77C56F1A3C@GVW1098EXB.americas.hpqcorp.net>
References: <A646510FAEBF484C81F1656322F311DC77C56F1A3C@GVW1098EXB.americas.hpqcorp.net>
Message-ID: <181622.41787.qm@web23702.mail.ird.yahoo.com>

I though ovS would support GRE directly as a virtual stack feature between OvS instances too.




----- Message d'origine ----
De : "Lee, Jeongkeun" <jklee at hp.com>
? : "discuss at openvswitch.org" <discuss at openvswitch.org>
Envoy? le : Jeu 25 mars 2010, 9 h 10 min 54 s
Objet : [ovs-discuss] GRE tunnel with 0.99.2 snapshot

Hi,

Does the current snapshot support GRE tunnel?
Based on the mailing list logs, gre support is under development and will be a part of the next release.

I just tried to attach a gre interface (created by linux 'ip') by ovs-vsctl add-port and ovs-dpctl add-if but none of them worked. I wonder there is any APIs used for testing.

I found a mention about gre interface type in the ovs-vswichtd.conf.db man page, though..

thanks,
-- JK

Jeongkeun "JK" Lee
Hewlett-Packard Labs
Palo Alto, CA



_______________________________________________
discuss mailing list
discuss at openvswitch.org
http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org



      


From blp at nicira.com  Thu Mar 25 12:48:35 2010
From: blp at nicira.com (Ben Pfaff)
Date: Thu, 25 Mar 2010 09:48:35 -0700
Subject: [ovs-discuss] brcompatd: "could not find valid configuration to
 update" error
In-Reply-To: <A646510FAEBF484C81F1656322F311DC77C56F19DE@GVW1098EXB.americas.hpqcorp.net>
References: <A646510FAEBF484C81F1656322F311DC77C56F18AA@GVW1098EXB.americas.hpqcorp.net>
	<20100324214413.GF7368@nicira.com>
	<A646510FAEBF484C81F1656322F311DC77C56F19DE@GVW1098EXB.americas.hpqcorp.net>
Message-ID: <20100325164835.GA30208@nicira.com>

You're right.

I pushed out a fix.

On Thu, Mar 25, 2010 at 02:51:44AM +0000, Lee, Jeongkeun wrote:
> Ben,
> 
> Thanks for the quick answers.
> I think INSTALL.bridge also needs to be fixed to use /usr/local.
> 
> thanks,
> JK
> 
> > -----Original Message-----
> > From: Ben Pfaff [mailto:blp at nicira.com]
> > Sent: Wednesday, March 24, 2010 2:44 PM
> > To: Lee, Jeongkeun
> > Cc: discuss at openvswitch.org
> > Subject: Re: [ovs-discuss] brcompatd: "could not find valid configuration
> > to update" error
> > 
> > On Wed, Mar 24, 2010 at 09:27:46PM +0000, Lee, Jeongkeun wrote:
> > > Today, I downloaded the latest snapshot from git and followed the
> > > instructions of INSTALL.bridge and INSTALL.Linux to use ovs as a
> > > replacement of Linux bridge.  OVS version is 0.99.2 and kernel is
> > > 2.6.26-2-amd64 on Debian lenny 5.03.
> > >
> > > Below are the commands that I ran after 'make install'
> > > completed. ovsdb-server, ovs-vswitchd, and ovs-brcompatd were called
> > > in separate shells to see the messages.
> > >
> > > % insmod datapath/linux-2.6/openvswitch_mod.ko
> > > % insmod datapath/linux-2.6/brcompat_mod.ko
> > > % ovsdb-tool create /etc/ovs-vswitchd.conf.db
> > vswitchd/vswitch.ovsschema
> > > % ovsdb-server /etc/ovs-vswitchd.conf.db --remote=punix:/var/run/ovsdb-
> > server
> > > % ovs-vswitchd -vANY:console:WARN unix:/var/run/ovsdb-server
> > > % ovs-brcompatd -vANY:console:WARN unix:/var/run/ovsdb-server
> > >
> > > ovsdb-server shows messages saying ovs-vswitchd and ovs-brcompatd are
> > connected
> > > and ovs-brcompatd complains
> > > 	database does not contain any Open vSwitch configuration
> > > when it gets started.
> > 
> > Thanks for reporting this problem.
> > 
> > You're right, there's a missing step here.  I'm going to add this
> > paragraph to INSTALL.Linux:
> > 
> >     Then initialize the database with "ovs-vsctl init".  This is only
> >     necessary the first time after you create the database with
> >     ovsdb-tool (but running it at any time is harmless):
> > 
> >           % ovs-vsctl init
> > 
> > after starting ovsdb-server and before starting ovs-vswitchd.
> > 
> > > Btw, ovs-vsctl connects to /usr/local/var/run/ovsdb-server by default
> > > instead of /var/run/ovsdb-server. Either ovs-vsctl or the INSTALL
> > > instruction needs to be revised to remove the discrepancy.
> > 
> > You are right again, of course.  The documentation is inconsistent here.
> > It assumes / in some places and /usr/local in others.  I will also push
> > out this fix:
> > 
> >     diff --git a/INSTALL.Linux b/INSTALL.Linux
> >     index 31bc3e0..263e261 100644
> >     --- a/INSTALL.Linux
> >     +++ b/INSTALL.Linux
> >     @@ -157,7 +157,7 @@ Prerequisites section, follow the procedure below
> > to build.
> > 
> >      7. Initialize the configuration database using ovsdb-tool, e.g.:
> > 
> >     -      % ovsdb-tool create /etc/ovs-vswitchd.conf.db
> > vswitchd/vswitch.ovsschema
> >     +      % ovsdb-tool create /usr/local/etc/ovs-vswitchd.conf.db
> > vswitchd/vswitch.ovsschema
> > 
> >      Startup
> >      =======
> >     @@ -167,7 +167,7 @@ configuration database, ovsdb-server.  Configure
> > it to use the
> >      database you created during step 7 of installation, above, and to
> >      listen on a Unix domain socket, e.g.:
> > 
> >     -      % ovsdb-server /etc/ovs-vswitchd.conf.db --
> > remote=punix:/var/run/ovsdb-server
> >     +      % ovsdb-server /usr/local/etc/ovs-vswitchd.conf.db --
> > remote=punix:/usr/local/var/run/ovsdb-server
> > 
> >      Then initialize the database with "ovs-vsctl init".  This is only
> >      necessary the first time after you create the database with
> >     @@ -178,7 +178,7 @@ ovsdb-tool (but running it at any time is
> > harmless):
> >      Then start the main Open vSwitch daemon, telling it to connect to
> > the
> >      same Unix domain socket:
> > 
> >     -      % ovs-vswitchd unix:/var/run/ovsdb-server
> >     +      % ovs-vswitchd unix:/usr/local/var/run/ovsdb-server
> > 
> >      Now you may use ovs-vsctl to set up bridges and other Open vSwitch
> >      features.  For example, to create a bridge named br0 and add ports
> > 
> > INSTALL.Linux does explain how to change the default:
> > 
> >    By default all files are installed under /usr/local.  If you want
> >    to install into, e.g., /usr and /var instead of /usr/local and
> >    /usr/local/var, add options as shown here:
> > 
> >       % ./configure --prefix=/usr --localstatedir=/var


From jpettit at nicira.com  Sat Mar 27 22:06:44 2010
From: jpettit at nicira.com (Justin Pettit)
Date: Sat, 27 Mar 2010 19:06:44 -0700
Subject: [ovs-discuss] OVS on VirtualBox?
In-Reply-To: <v2kef6fabde1003271843mb8b6808albda6edcee9f83cf@mail.gmail.com>
References: <v2kef6fabde1003271843mb8b6808albda6edcee9f83cf@mail.gmail.com>
Message-ID: <D700C83A-F596-4B64-86C5-B7508AE510A4@nicira.com>

Hi, Guilherme.  I have not personally run OVS with VirtualBox, but we've had reports that it works with Linux as the host.  Doing a quick pass through their documentation, it appears that you should be able to make it work simply by running OVS with bridge compatibility.  This is documented in INSTALL.bridge in the distribution.  You can also browse the documentation here:

	http://openvswitch.org/cgi-bin/gitweb.cgi?p=openvswitch;a=blob_plain;f=INSTALL.bridge

If you run into any problems, feel free to send questions to this list.  If you get it working, we'd love to hear about that, too!

--Justin


On Mar 27, 2010, at 6:43 PM, Guilherme Fernandes wrote:

> Hi all,
> 
> I'm interested in creating a virtual dev environment with several pcs and a manageable (virtual) switch with VLAN support. I already use VirtualBox for some of the machines, so I would like to add OVS into the mix.
> 
> I've seen a few pages mentioning OVS supports open source VirtualBox, but I can find no documentation at all about it. Is the mentioned support being that it can be used as a drop-in for VirtualBox's vswitch? If so, is this accomplished by just installing OVS on the Linux host where VirtualBox will run? Can someone elaborate a little on how they integrate?
> 
> Thanks a lot in advance,
> Guilherme
> _______________________________________________
> discuss mailing list
> discuss at openvswitch.org
> http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org



From jpettit at nicira.com  Sat Mar 27 22:43:44 2010
From: jpettit at nicira.com (Justin Pettit)
Date: Sat, 27 Mar 2010 19:43:44 -0700
Subject: [ovs-discuss] OVS on VirtualBox?
In-Reply-To: <t2pef6fabde1003271915q39233978jf71055491e1e094e@mail.gmail.com>
References: <v2kef6fabde1003271843mb8b6808albda6edcee9f83cf@mail.gmail.com>
	<D700C83A-F596-4B64-86C5-B7508AE510A4@nicira.com>
	<t2pef6fabde1003271915q39233978jf71055491e1e094e@mail.gmail.com>
Message-ID: <0C9B3474-90BC-4FA0-B2DC-5B1D26CD8A64@nicira.com>

Unfortunately, I haven't heard much details about how people were using it.  From a quick glance at the documentation, I didn't see exactly how Internal and Host Networking is done.  However, if these are simply additional bridges with special devices attached (or not in the case of Host Networking), then it should "just work" with bridge compatibility.  I'd give it a shot, and if it doesn't work, let us know.  I can't promise that we can devote many resources to fixing problems specific to VirtualBox, but at the very least we can do some basic troubleshooting.  And of course, we always take patches and better documentation!

As an aside, I tried to get VirtualBox running on my Mac a few months ago, but it seemed to dislike VMware fusion, and caused a kernel panic every time.  I should really get it running natively on a Linux box, but other hypervisors have been a higher priority lately.

--Justin


On Mar 27, 2010, at 7:15 PM, Guilherme Fernandes wrote:

> Hi Justin,
> 
> Thanks a lot for the quick response! I'll try it out and report back. 
> 
> I would still like to hear more details from the people that tried it out (e.g. does VBox Internal/Host Networking modes also work, etc), but this should get me started.
> 
> Thanks,
> Guilherme
> 
> On Sat, Mar 27, 2010 at 10:06 PM, Justin Pettit <jpettit at nicira.com> wrote:
> Hi, Guilherme.  I have not personally run OVS with VirtualBox, but we've had reports that it works with Linux as the host.  Doing a quick pass through their documentation, it appears that you should be able to make it work simply by running OVS with bridge compatibility.  This is documented in INSTALL.bridge in the distribution.  You can also browse the documentation here:
> 
>        http://openvswitch.org/cgi-bin/gitweb.cgi?p=openvswitch;a=blob_plain;f=INSTALL.bridge
> 
> If you run into any problems, feel free to send questions to this list.  If you get it working, we'd love to hear about that, too!
> 
> --Justin
> 
> 
> On Mar 27, 2010, at 6:43 PM, Guilherme Fernandes wrote:
> 
> > Hi all,
> >
> > I'm interested in creating a virtual dev environment with several pcs and a manageable (virtual) switch with VLAN support. I already use VirtualBox for some of the machines, so I would like to add OVS into the mix.
> >
> > I've seen a few pages mentioning OVS supports open source VirtualBox, but I can find no documentation at all about it. Is the mentioned support being that it can be used as a drop-in for VirtualBox's vswitch? If so, is this accomplished by just installing OVS on the Linux host where VirtualBox will run? Can someone elaborate a little on how they integrate?
> >
> > Thanks a lot in advance,
> > Guilherme
> > _______________________________________________
> > discuss mailing list
> > discuss at openvswitch.org
> > http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org
> 
> 



From amin at cs.toronto.edu  Sun Mar 28 02:08:28 2010
From: amin at cs.toronto.edu (Amin Tootoonchian)
Date: Sun, 28 Mar 2010 02:08:28 -0400
Subject: [ovs-discuss] [Open vSwitch 0.99.2] Kernel panic with eth0 in the
	bridge
Message-ID: <205db5d41003272308g47f5cba5hcf0a638f4e217843@mail.gmail.com>

Hi,

openvswitch_mod kernel module causes a kernel panic when it adds eth0
to the bridge. It seems to run fine with all the interfaces except
eth0 (which is connected to a switch). I verified that I have no
problems with Linux bridging with the same set of interfaces. Do you
know what is wrong?


Thanks,
Amin

== OS
Debian Lenny 2.6.26-2-openvz-686 (32 bit)

== ovs.conf
bridge.mybr.port=eth0
bridge.mybr.port=veth31100.0
bridge.mybr.port=veth31101.0

== kernel panic
Mar 28 01:32:59|00001|cfg|INFO|using "ovs.conf" as configuration file,
".ovs.conf.~lock~" as lock file
[  205.739294] device mybr entered promiscuous mode
Mar 28 01:32:59|00002|ofproto|INFO|using datapat[  205.799984] device
eth0 entered promiscuous mode
h ID 002320e2fe55
Mar 28 01:32:59|00003|bridge|[  205.828857] device veth31100.0 entered
promiscuous mode
[  205.839293] device veth31101.0 entered promiscuous mode

Mar 28 01:32:59|00004|bridge|INFO|created port eth0 on bridge mybr
Mar 28 01:32:59|00005|bridge|INFO|created port veth31100.0 on bridge mybr
Mar 28 01:32:59|00006|bridge|INFO|created port veth31101.0 on bridge mybr
Mar 28 01:32:59|00007|ofproto|INFO|datapath ID changed to 001851e34ae8
[  205.949500] BUG: unable to handle kernel NULL pointer dereference at 00000000
[  205.967857] IP: [<00000000>]
[  205.980949] *pdpt = 0000000026074001 *pde = 0000000000000000
[  205.996952] Oops: 0000 [#1] SMP
[  205.999713] Modules linked in: brcompat_mod openvswitch_mod
vzethdev vznetdev simfs vzrst vzcpt tun vzdquota vzmon ipv6 vzdev
xt_tcpudp xt_length ipt_ttl xt_tcpmss xt_TCPMSS iptable_mangle
iptable_filter xt_multiport xt_limit xt_dscp ipt_REJECT ip_tables
x_tables nfs lockd nfs_acl sunrpc psmouse ipmi_si hpilo pcspkr
ipmi_msghandler serio_raw container shpchp button pci_hotplug evdev
ext3 jbd mbcache sg sr_mod cdrom sd_mod usbhid hid ff_memless ata_piix
ata_generic libata scsi_mod dock ide_pci_generic uhci_hcd ehci_hcd
ide_core usbcore tg3 e1000e thermal processor fan thermal_sys [last
unloaded: scsi_wait_scan]
[  206.191244]
[  206.191244] Pid: 3733, comm: ovs-vswitchd Not tainted
(2.6.26-2-openvz-686 #1 036test001)
[  206.191244] EIP: 0060:[<00000000>] EFLAGS: 00010297 CPU: 3
[  206.191244] EIP is at 0x0
[  206.191244] EAX: e7935b40 EBX: e5036800 ECX: 4baeea0b EDX: e7992a40
[  206.191244] ESI: e7935b40 EDI: c03888e8 EBP: e5036800 ESP: e58bbe40
[  206.191244]  DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
[  206.191244] Process ovs-vswitchd (pid: 3733, veid: 0, ti=e58ba000
task=e752eb50 task.ti=e58ba000)
[  206.191244] Stack: c0260ae6 e5036800 00000003 e7935b40 e6000008
c0263244 c0166cdd 0000003c
[  206.191244]        e7935b40 000005dc e6000008 e95fd1ce 0000002a
e50a4000 e58bbeec 00000002
[  206.191244]        e6925b00 e6001aa0 e95fd340 e58bbeec e6925b00
e50a4000 e5409fa0 00000282
[  206.191244]  Call Trace:
[  206.191244]  [<c0260ae6>] dev_hard_start_xmit+0x1fd/0x286
[  206.191244]  [<c0263244>] dev_queue_xmit+0x211/0x2c5
[  206.191244]  [<c0166cdd>] __alloc_pages_internal+0x3ac/0x3c7
[  206.191244]  [<e95fd1ce>] dp_xmit_skb+0x79/0x82 [openvswitch_mod]
[  206.191244]  [<e95fd340>] execute_actions+0x128/0x6fc [openvswitch_mod]
[  206.191244]  [<c025b852>] __alloc_skb+0x5c/0x120
[  206.191244]  [<e95ff376>] openvswitch_ioctl+0xd08/0xd63 [openvswitch_mod]
[  206.191244]  [<c011b424>] do_page_fault+0x515/0x8cd
[  206.191244]  [<c02c5cfa>] error_code+0x72/0x78
[  206.191244]  [<e95fe66e>] openvswitch_ioctl+0x0/0xd63 [openvswitch_mod]
[  206.191244]  [<c018ecf8>] vfs_ioctl+0x1c/0x5d
[  206.191244]  [<c018ef83>] do_vfs_ioctl+0x24a/0x261
[  206.191244]  [<c01e2d5c>] __copy_to_user_ll+0x21/0x2c
[  206.191244]  [<c018efc6>] sys_ioctl+0x2c/0x42
[  206.191244]  [<c0108976>] syscall_call+0x7/0xb
[  206.191244]  [<c02c0000>] init_amd+0x11c/0x66e
[  206.191244]  =======================
[  206.191244] Code:  Bad EIP value.
[  206.191244] EIP: [<00000000>] 0x0 SS:ESP 0068:e58bbe40

Message from syslogd at xyz at Mar 28 01:33[  207.656913] Kernel panic -
not syncing: Fatal exception in interrupt


From amin at cs.toronto.edu  Mon Mar 29 11:18:58 2010
From: amin at cs.toronto.edu (Amin Tootoonchian)
Date: Mon, 29 Mar 2010 11:18:58 -0400
Subject: [ovs-discuss] [Open vSwitch 0.99.2] Kernel panic with eth0 in
	the bridge
In-Reply-To: <25cfcaf91003290724q2bd93f81ob9a1a91c9bbbbbfb@mail.gmail.com>
References: <205db5d41003272308g47f5cba5hcf0a638f4e217843@mail.gmail.com> 
	<25cfcaf91003290724q2bd93f81ob9a1a91c9bbbbbfb@mail.gmail.com>
Message-ID: <205db5d41003290818x10abcc0eg4b7108409b4a567@mail.gmail.com>

Thanks Jesse. Do have any plans to add support for OpenVZ kernels?

Amin

On Mon, Mar 29, 2010 at 10:24 AM, Jesse Gross <jesse at nicira.com> wrote:
> On Sun, Mar 28, 2010 at 2:08 AM, Amin Tootoonchian <amin at cs.toronto.edu>
> wrote:
>>
>> Hi,
>>
>> openvswitch_mod kernel module causes a kernel panic when it adds eth0
>> to the bridge. It seems to run fine with all the interfaces except
>> eth0 (which is connected to a switch). I verified that I have no
>> problems with Linux bridging with the same set of interfaces. Do you
>> know what is wrong?
>
> OpenVZ adds a couple of extensions to the networking stack, including a new
> bridge hook on transmit. ?The version of the bridge in the OpenVZ kernel
> knows how to handle these extensions, which is why it works fine. ?Open
> vSwitch doesn't support OpenVZ and doesn't set the transmit bridge hook
> (probably other things too), which is why you get a null pointer deference.


From blp at nicira.com  Tue Mar 30 12:22:38 2010
From: blp at nicira.com (Ben Pfaff)
Date: Tue, 30 Mar 2010 09:22:38 -0700
Subject: [ovs-discuss] dump-flow doesn't show default flow
In-Reply-To: <8CC9E21CAB3595C-1E5C-19509@webmail-d091.sysops.aol.com>
References: <8CC9E21CAB3595C-1E5C-19509@webmail-d091.sysops.aol.com>
Message-ID: <20100330162238.GB11093@nicira.com>

On Tue, Mar 30, 2010 at 09:12:32AM -0400, devang.vyas at aol.in wrote:
> I am using version 0.99.1 of open vswitch on my Xen Server
> 5.5. ovs-ofctl dump-flow command doesn't show default flow. And also
> other flows doesn't match with traffic.

Are you using a controller?  If you are not, then you should see one
flow, e.g.:

    [root at localhost ~]# ovs-ofctl dump-flows xenbr0
    Mar 30 09:18:55|00001|ofctl|INFO|connecting to unix:/var/run/openvswitch/xenbr0.mgmt
    stats_reply (xid=0x73515d7e): flags=none type=1(flow)
      cookie=0, duration_sec=68016s, duration_nsec=905000000ns, table_id=1, priority=0, n_packets=414970, n_bytes=46119909, actions=NORMAL

You will not see any other flows, because this flow handles all traffic.

If you are using a controller, then there are no default flows.

> This problem occurred after install/uninstall rpm repeateadly few
> times. 

You characterize this as a problem.  Does that mean that something is
not working correctly?

> Is there any clean up procedure required before installing new rpm?

No.