Re: [ovs-dev] [PATCH] ovn: make ipam tests more reliable

2016-09-14 Thread Ben Pfaff
On Wed, Sep 14, 2016 at 08:23:44PM -0400, Lance Richardson wrote:
> > From: "Ben Pfaff" 
> > To: "Lance Richardson" 
> > Cc: dev@openvswitch.org
> > Sent: Wednesday, September 14, 2016 8:03:16 PM
> > Subject: Re: [ovs-dev] [PATCH] ovn: make ipam tests more reliable
> > 
> > On Wed, Sep 14, 2016 at 07:30:50PM -0400, Lance Richardson wrote:
> > > After adding log messages to better understand  IPAM-related code
> > > in ovn northd, the IPAM tests began to fail occasionally. Adding
> > > --wait=sb to commands triggering address allocation eliminated
> > > these failures (there were no failures with 100 executions when
> > > testing with this change).
> > > 
> > > Signed-off-by: Lance Richardson 
> > 
> > I think that this adds more "--wait"s than strictly necessary (it's only
> > really useful on the last ovn-nbctl call in a series of them) but that's
> > not really harmful (except to performance), so I applied it to master
> > and branch-2.6.
> > 
> I did wonder whether it might be worthwhile to avoid the --wait
> in each iteration of the loop and just have one --wait after the loop, but
> I wasn't sure about the best way to go about doing that.

One option is to add a separate "ovn-nbctl --wait=sb sync" command after
the changes.

> This was the first time I had tried using the --wait option, it's a nice
> addition and much better than trying to band-aid with sleeps.

I am not sure that it is completely fail-safe at this point, but I think
it is a big step forward.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev


Re: [ovs-dev] [PATCH] ovn: make ipam tests more reliable

2016-09-14 Thread Lance Richardson
> From: "Ben Pfaff" 
> To: "Lance Richardson" 
> Cc: dev@openvswitch.org
> Sent: Wednesday, September 14, 2016 8:03:16 PM
> Subject: Re: [ovs-dev] [PATCH] ovn: make ipam tests more reliable
> 
> On Wed, Sep 14, 2016 at 07:30:50PM -0400, Lance Richardson wrote:
> > After adding log messages to better understand  IPAM-related code
> > in ovn northd, the IPAM tests began to fail occasionally. Adding
> > --wait=sb to commands triggering address allocation eliminated
> > these failures (there were no failures with 100 executions when
> > testing with this change).
> > 
> > Signed-off-by: Lance Richardson 
> 
> I think that this adds more "--wait"s than strictly necessary (it's only
> really useful on the last ovn-nbctl call in a series of them) but that's
> not really harmful (except to performance), so I applied it to master
> and branch-2.6.
> 
I did wonder whether it might be worthwhile to avoid the --wait
in each iteration of the loop and just have one --wait after the loop, but
I wasn't sure about the best way to go about doing that.

This was the first time I had tried using the --wait option, it's a nice
addition and much better than trying to band-aid with sleeps.

Thanks,

   Lance

> Thank you!
> 
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev


Re: [ovs-dev] [PATCH] ovn: make ipam tests more reliable

2016-09-14 Thread Ben Pfaff
On Wed, Sep 14, 2016 at 07:30:50PM -0400, Lance Richardson wrote:
> After adding log messages to better understand  IPAM-related code
> in ovn northd, the IPAM tests began to fail occasionally. Adding
> --wait=sb to commands triggering address allocation eliminated
> these failures (there were no failures with 100 executions when
> testing with this change).
> 
> Signed-off-by: Lance Richardson 

I think that this adds more "--wait"s than strictly necessary (it's only
really useful on the last ovn-nbctl call in a series of them) but that's
not really harmful (except to performance), so I applied it to master
and branch-2.6.

Thank you!
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev