[c-nsp] PFR Question
David Prall
dcp at dcptech.com
Sat Mar 27 14:49:13 EDT 2010
PfR takes care of the rerouting on a site basis. The site is monitoring
reachability to a particular prefix. The key issue with a single cloud, is
that you don't control the end to end path. If it is two clouds then you can
monitor end to end via each cloud, and choose which one is better to use for
a particular prefix or traffic type.
--
http://dcp.dcptech.com
> -----Original Message-----
> From: jack daniels [mailto:jckdaniels12 at gmail.com]
> Sent: Friday, March 26, 2010 10:20 PM
> To: David Prall
> Cc: cisco-nsp at puck.nether.net
> Subject: Re: [c-nsp] PFR Question
>
> IN SCENARIO BOTH LINKS FROM SAME SERVICE PROVIDER -But how will this
> avoid drops when PE1and CE1 link goes down as MPBGP bring secondary
> path as best in BGP table ( MPLS domain )and then to routing table will
> take atleast 3 min.
> Till secondry path not in routing table there will be pcket drops.So
> PE3 will converge so fast.
>
>
>
> On 3/26/10, David Prall <dcp at dcptech.com> wrote:
>
> This is where PfR is involved to route around the primary carrier
> to the
> secondary.
>
> --
> http://dcp.dcptech.com
>
> > -----Original Message-----
> > From: jack daniels [mailto:jckdaniels12 at gmail.com]
> > Sent: Thursday, March 25, 2010 8:50 PM
> > To: David Prall
> > Cc: cisco-nsp at puck.nether.net
> > Subject: Re: [c-nsp] PFR Question
> >
> > Hi David,
> >
> > In a multipath instance PE1 will install the Equal Cost route
> with rd
> > 1:1
> > first, using 1:2 as a secondary path only. Opposite on PE2.???
> > whne both paths have equal cost the why route with rd1:1 will
> be
> > primary always
> > and rd 1:2 will be secondary on PE1.
> >
> > EVEN IF WE advertise X.X.X.X from PE1 and PE2 still PE3 will
> have two
> > routes in BGP table . But one in routing table.
> > But how will this avoid drops when PE1and CE1 link goes down as
> BGP
> > bring secondary path to Primary and then to routing table will
> take
> > atleast 3 min.
> >
> > Regards
> >
> >
> >
> > On Fri, Mar 26, 2010 at 12:29 AM, David Prall <dcp at dcptech.com>
> wrote:
> >
> >
> > 1)
> > On PE1
> > vrf description customer
> > rd 1:1
> > route-target both 1:1
> > route-target import 1:2
> > On PE2
> > vrf description customer
> > rd 1:2
> > route-target both 1:2
> > route-target import 1:1
> >
> > In a multipath instance PE1 will install the Equal Cost
> route
> > with rd 1:1
> > first, using 1:2 as a secondary path only. Opposite on
> PE2.
> >
> > 2)
> > Could use different VRF's. Just like dual carriers. A key
> concern
> > is a dual
> > failure, site 1 on network 1 and site 2 on network 2. The
> > customer will need
> > to provide a path between the two networks via one of
> their
> > sites.
> >
> >
> > David
> >
> > --
> > http://dcp.dcptech.com <http://dcp.dcptech.com/>
> >
> >
> > > -----Original Message-----
> > > From: jack daniels [mailto:jckdaniels12 at gmail.com]
> >
> > > Sent: Thursday, March 25, 2010 2:41 PM
> > > To: David Prall
> > > Cc: cisco-nsp at puck.nether.net
> > > Subject: Re: [c-nsp] PFR Question
> > >
> >
> > > Hi David ,
> > >
> > > thanks man I got the basic idea :)
> > >
> > > 1) but please explain in more detail this
> > >
> > > Single VRF, 2 distinct RD's. The VRF imports both,
> exports one.
> > The
> > > RD's are
> > > different so that multipath can be used within the core
> > typically. But
> > > in
> > > this case they wouldn't use multipath and the local RD
> would be
> > used as
> > > the
> > > determining factor on import of which route is
> installed
> > first.??????
> > >
> > >
> > > 2) Also if I use diffrent VRF for CE4---CE2 path that
> will also
> > work -
> > > ??
> > >
> > >
> > > On Thu, Mar 25, 2010 at 11:57 PM, David Prall
> <dcp at dcptech.com>
> > wrote:
> > >
> > >
> > > If the link goes away, then the update should be
> pretty
> > quick.
> > >
> > > Single VRF, 2 distinct RD's. The VRF imports
> both,
> > exports one.
> > > The RD's are
> > > different so that multipath can be used within
> the core
> > > typically. But in
> > > this case they wouldn't use multipath and the
> local RD
> > would be
> > > used as the
> > > determining factor on import of which route is
> installed
> > first.
> > >
> > > The local CE (CE3) is probing for the subnet at
> CE1. When
> > it is
> > > no longer
> > > reachable by CE3 it will move the route to CE4.
> As long
> > as CE4 is
> > > using CE2
> > > as the path via the cloud then no issue.
> > >
> > >
> > > David
> > >
> > > --
> >
> > > http://dcp.dcptech.com <http://dcp.dcptech.com/>
> > <http://dcp.dcptech.com/>
> >
> > >
> > >
> > > > -----Original Message-----
> > > > From: jack daniels
> [mailto:jckdaniels12 at gmail.com]
> > >
> > > > Sent: Thursday, March 25, 2010 2:19 PM
> > > > To: David Prall
> > > > Cc: cisco-nsp at puck.nether.net
> > > > Subject: Re: [c-nsp] PFR Question
> > > >
> > >
> > > > If a single carrier, then the CE4/CE2 path
> needs to be
> > via
> > > > a second RD so that the paths within the
> carrier are
> > preferred
> > > and the
> > > > same
> > > > will happen.????
> > > > DO YOU mean we need to have diifrent vrf on
> secondry
> > end to end
> > > path.
> > > >
> > > > I didnt get this if single carrier as link PE1
> and CE1
> > link
> > > fails
> > > > ....CE3 send traffic for X.X.X.X to PE3.PE3
> still has
> > next hop
> > > in its
> > > > vrf table as PE1....
> > > >
> > > >
> > > >
> > > >
> > > > Please help me as still confused if two
> carriers , how
> > will
> > > this
> > > > hhappen
> > > >
> > > >
> > > > On Thu, Mar 25, 2010 at 11:29 PM, David Prall
> > <dcp at dcptech.com>
> > > wrote:
> > > >
> > > >
> > > > Is MPLS Domain a single carrier, or two
> carriers.
> > If two
> > > carriers
> > > > then the
> > > > CE3/CE4 site will see that they can't
> reach via
> > CE3/CE1
> > > path and
> > > > switch over
> > > > to CE4/CE2 path. If a single carrier,
> then the
> > CE4/CE2
> > > path needs
> > > > to be via
> > > > a second RD so that the paths within the
> carrier
> > are
> > > preferred
> > > > and the same
> > > > will happen. PfR is providing end-to-end
> > reachability
> > > information
> > > > in this
> > > > case, and based on that changing the
> local
> > routing table.
> > > >
> > > > David
> > > >
> > > >
> > > > --
> > >
> > > > http://dcp.dcptech.com
> <http://dcp.dcptech.com/>
> > <http://dcp.dcptech.com/>
> > > <http://dcp.dcptech.com/>
> > >
> > > >
> > > >
> > > > > -----Original Message-----
> > > >
> > > > > From: jack daniels
> > [mailto:jckdaniels12 at gmail.com]
> > > > > Sent: Thursday, March 25, 2010 1:07 PM
> > > > > To: David Prall
> > > > > Cc: cisco-nsp at puck.nether.net
> > > > > Subject: Re: [c-nsp] PFR Question
> > > > >
> > > >
> > > > > But if you have --
> > > > >
> > > > >
> > > > > |CE1--------PE1
> > > > > PE3--------CE3
> > > > > X.X.X.X---------| ----
> --------
> > --------
> > > MPLS
> > > > DOMAIN-----
> > > > > --------------
> > > > > | CE2--------PE2
> > > > > PE4--------CE4
> > > > >
> > > > >
> > > > > Now my primary link is CE1-PE1 and
> secondary is
> > CE2-PE2
> > > > > If my CE1-PE1 goes down i route traffic
> via
> > CE2-
> > > PE2<<<<<<I
> > > > understand
> > > > > this ok...
> > > > >
> > > > > when traffic from CE3 for X.X.X.X
> reaches PE3 ,
> > next
> > > hop is
> > > > still PE1 (
> > > > > as MPBGP has not converged so fast in
> MPLS
> > domain of
> > > SP) ...so
> > > > how will
> > > > > traffic be forwareded , as PFR claims 3
> sec.
> > > > >
> > > > >
> > > > > On Thu, Mar 25, 2010 at 10:16 PM, David
> Prall
> > > <dcp at dcptech.com>
> > > > wrote:
> > > > >
> > > > >
> > > > > PfR is a unidirectional feature.
> The
> > router on
> > > the other
> > > > end
> > > > > needs to be
> > > > > configured with PfR as well in
> order to
> > have
> > > > bidirectional
> > > > > visibility.
> > > > > Typically the master controller
> will be
> > local to
> > > the
> > > > site.
> > > > >
> > > > > --
> > > >
> > > > > http://dcp.dcptech.com
> > <http://dcp.dcptech.com/> <http://dcp.dcptech.com/>
> > > <http://dcp.dcptech.com/>
> > > > <http://dcp.dcptech.com/>
> > > >
> > > > >
> > > > >
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: cisco-nsp-
> bounces at puck.nether.net
> > > [mailto:cisco-
> > > > nsp-
> > > > > > bounces at puck.nether.net] On
> Behalf Of
> > jack
> > > daniels
> > > > > > Sent: Thursday, March 25, 2010
> 12:35 PM
> > > > > > To: cisco-nsp at puck.nether.net
> > > > > > Subject: Re: [c-nsp] PFR
> Question
> > > > > >
> > > > > > dear guys,
> > > > > >
> > > > > > is my mail being delivered to
> group as
> > no one
> > > replied.
> > > > > >
> > > > > > On Wed, Mar 24, 2010 at 11:42
> PM, jack
> > daniels
> > > > > > <jckdaniels12 at gmail.com>wrote:
> > > > > >
> > > > > > > Hi Network champs,
> > > > > > >
> > > > > > > I'm stuck in understanding of
> PFR .
> > Docs say
> > > it
> > > > converges in
> > > > > 3 sec (
> > > > > > for
> > > > > > > realtime traffic VOICE )...
> > > > > > >
> > > > > > > I understand you can send
> traffic out
> > > secondry link
> > > > but what
> > > > > about
> > > > > > traffic
> > > > > > > which has to come back from
> remote
> > end ( for
> > > which SP
> > > > has not
> > > > > > converged).
> > > > > > >
> > > > > > > But if you have --
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > |CE1--------PE1
> > > > > > > PE3--------CE3
> > > > > > > X.X.X.X---------|
> --
> > --------
> > > --------
> > > > --MPLS
> > > > > > > DOMAIN-------------------
> > > > > > > | CE2--
> ------
> > PE2
> > > > > > > PE4--------CE4
> > > > > > >
> > > > > > >
> > > > > > > Now my primary link is CE1-
> PE1 and
> > secondary
> > > is CE2-
> > > > PE2
> > > > > > > If my CE1-PE1 goes down i
> route
> > traffic via
> > > CE2-
> > > > PE2<<<<<<I
> > > > > understand
> > > > > > this
> > > > > > > ok...
> > > > > > >
> > > > > > >
> > > > > > > BUT MY QUESTION IS -
> > > > > > >
> > > > > > > PE3 and PE4 ( for this VRF)
> still has
> > NOW
> > > converged
> > > > the BGP
> > > > > and still
> > > > > > for
> > > > > > > it next hop for X.X.X.X is
> PE1. So
> > how fwd
> > > can happen
> > > > in 3
> > > > > sec untill
> > > > > > > Service providers all routers
> dont
> > converge
> > > and
> > > > understand
> > > > > that CE1-
> > > > > > PE1 link
> > > > > > > is down.
> > > > > > >
> > > > > > >
> > > > > > > Regards
> > > > > > >
> > > > > > >
> > > > >
> > > > > >
> > _______________________________________________
> > > > > > cisco-nsp mailing list cisco-
> > > nsp at puck.nether.net
> > > > > >
> > https://puck.nether.net/mailman/listinfo/cisco-
> > > nsp
> > > > > > archive at
> > > http://puck.nether.net/pipermail/cisco-nsp/
> > > > >
> > > > >
> > > > >
> > > >
> > > >
> > > >
> > > >
> > >
> > >
> > >
> > >
> >
> >
> >
> >
>
>
>
>
More information about the cisco-nsp
mailing list