[ovs-discuss] [PATCH] ovs-brcompatd: Document bug. (was: Re: duplicate port names in DB)

Ben Pfaff blp at nicira.com
Tue May 3 17:33:34 UTC 2011

On Tue, May 03, 2011 at 07:06:31PM +0200, Gregor Schaffrath wrote:
> Anyhow: the workaround is to start with an empty DB at bootup every
> time (maybe it'd make sense to point this out as a caveat somewhere
> until it's fixed, or the decision to remove it has been made :) )

It turns out that this is not a regression, it's a bug that
ovs-brcompatd has always had.  (We should still fix it, of course.)

For the moment, I think that, as you say, it's best to document it:

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

From: Ben Pfaff <blp at nicira.com>
Date: Tue, 3 May 2011 10:30:17 -0700
Subject: [PATCH] ovs-brcompatd: Document bug.

Reported-by: Gregor Schaffrath <grsch at net.t-labs.tu-berlin.de>
 AUTHORS                     |    1 +
 vswitchd/ovs-brcompatd.8.in |    6 ++++++
 2 files changed, 7 insertions(+), 0 deletions(-)

diff --git a/AUTHORS b/AUTHORS
index 38ca142..5bb0b4a 100644
@@ -55,6 +55,7 @@ DK Moon                 dkmoon at nicira.com
 Gaetano Catalli         gaetano.catalli at gmail.com
 Ghanem Bahri            bahri.ghanem at gmail.com
 Giuseppe de Candia      giuseppe.decandia at gmail.com
+Gregor Schaffrath       grsch at net.t-labs.tu-berlin.de
 Hassan Khan             hassan.khan at seecs.edu.pk
 Hector Oron             hector.oron at gmail.com
 Henrik Amren            henrik at nicira.com
diff --git a/vswitchd/ovs-brcompatd.8.in b/vswitchd/ovs-brcompatd.8.in
index 0cce84d..2ce164c 100644
--- a/vswitchd/ovs-brcompatd.8.in
+++ b/vswitchd/ovs-brcompatd.8.in
@@ -52,6 +52,12 @@ pruned.  The default prune timeout is 5 seconds.
 .so lib/common.man
 .so lib/leak-checker.man
+\fBovs\-brcompatd\fR requires the bridges that it manages to initially
+have no ports listed in their database records when it starts up.
+Otherwise, it may add duplicate ports to bridges.
 \fBovs\-brcompatd\fR requires the \fBbrcompat_mod.ko\fR kernel module to be

More information about the discuss mailing list