Thread: Re: [mpls-linux-general] ldp_linux don't work for [static] README.sample environment!
Status: Beta
Brought to you by:
jleu
Menu
▾
▴
mpls-linux-general
|
From: Markus H. <mh...@as...> - 2003-02-16 23:55:23
|
Hi all, any news about this? Looks like with CVS revision 1.16 (local p4 tree @274) its still not working... Any (simple) workaround? Thx in advance, Markus James R. Leu wrote: > The problem you are seeing (LSP setup in one direction) is a know > problem. I'm working on a fix in the zebra<->ldp-portable version > (grab it from CVS to see the progress thus far) > > Jim > > On Tue, Jun 11, 2002 at 01:20:42AM -0300, Morvan Daniel Müller wrote: > > Hello all! > > > > I want to make ldp-linux work for this simple testbed: > > > > My testbed configuration is: > > hostA LERA CORE LERB hostB > > 10.1.0.2/24----10.1.0.1/24 10.2.1.1/24----10.2.1.2/24 > > 10.1.1.2/24-----10.1.1.1/24 10.2.0.1/24----10.2.0.2/24 > > > > I'am using only host static routes to make hostA and hostB are > reacheable to each other. > > Static Routes added at: > > LERA: > > route add 10.2.1.2/32 gw 10.1.0.1 > > LERB: > > route add 10.1.1.2/32 gw 10.2.0.1 > > CORE: > > route add 10.1.1.2/32 gw 10.1.0.2 > > route add 10.2.1.2/32 gw 10.2.0.2 > > > > For tests I use the command "traceroute" with: > > IP_only (ip_forwarding) > > LDP: "mpls-adm" and "ldp-linux". > > > > I were Waiting this Results: > > traceroute -n from hostA to host B: may show 4 hops with standard IP, > an 3 hops with LDP (mpls_adm or ldp_linux). > > > > For IP (ip_forwarding only) it's ok - 4 hops. > > For LDP using "mpls_adm" it's ok - 3 hops > > For LDP using "ldp_linux" IT NEVER WORK!!!!!!!!!! > > > > I see that "ldp_linux" only create the LSP for one direction (from A > to B)! > > Like documentation "ldp_linux" create LSPs basead on the L3 > information (so, static routes created above should result in LSPs > created in two directions)! > > > > I follow README.sample commands, but for the sample environment it > too don't work! > > > > If some one in this mailing list had sucessfull configured > "ldp_linux" in such environment > > or like README.sample and the LSPs are dinamically create for two > directions, please let me know! > > > > Morvan. > > > > _______________________________________________________________ > > > > Don't miss the 2002 Sprint PCS Application Developer's Conference > > August 25-28 in Las Vegas - > http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink > > > > _______________________________________________ > > mpls-linux-general mailing list > > mpl...@li... > > https://lists.sourceforge.net/lists/listinfo/mpls-linux-general > > -- > James R. Leu > |
|
From: James R. L. <jl...@mi...> - 2003-02-18 04:02:46
|
On Mon, Feb 17, 2003 at 12:55:06AM +0100, Markus Hennig wrote: > Hi all, >=20 > any news about this? >=20 > Looks like with CVS revision 1.16 (local p4 tree @274) its still not=20 > working... >=20 > Any (simple) workaround? >=20 > Thx in advance, > Markus The title has me a bit confused. ldp_linux no longer exists, the only wa= y I provide for you to run LDP is with zebra. (using the mplsd daemon). As far as I know mplsd can create LSPs correctly for the intended environ= ment (LSPs being setup for the LDP Identifier) in DU ordered control. Of cour= se not every one understands the ramifications of that statment. Start out by describing what you are trying to do, what you are observing and what you expected to see. Please include a ascii drawing of you setu= p. >=20 > James R. Leu wrote: > > The problem you are seeing (LSP setup in one direction) is a know > > problem. I'm working on a fix in the zebra<->ldp-portable version > > (grab it from CVS to see the progress thus far) > >=20 > > Jim > >=20 > > On Tue, Jun 11, 2002 at 01:20:42AM -0300, Morvan Daniel M=FCller wrot= e: > > > Hello all! > > > > > > I want to make ldp-linux work for this simple testbed: > > > > > > My testbed configuration is: > > > hostA LERA CORE LERB hostB > > > 10.1.0.2/24----10.1.0.1/24 10.2.1.1/24----10.2.= 1.2/24 > > > 10.1.1.2/24-----10.1.1.1/24 10.2.0.1/24----10.2.0.2/24 =20 > > > > > > I'am using only host static routes to make hostA and hostB are=20 > > reacheable to each other. > > > Static Routes added at: > > > LERA: > > > route add 10.2.1.2/32 gw 10.1.0.1 > > > LERB: > > > route add 10.1.1.2/32 gw 10.2.0.1 > > > CORE: > > > route add 10.1.1.2/32 gw 10.1.0.2 > > > route add 10.2.1.2/32 gw 10.2.0.2 > > >=20 > > > For tests I use the command "traceroute" with: > > > IP_only (ip_forwarding) > > > LDP: "mpls-adm" and "ldp-linux". > > > > > > I were Waiting this Results: > > > traceroute -n from hostA to host B: may show 4 hops with standard = IP,=20 > > an 3 hops with LDP (mpls_adm or ldp_linux). > > > > > > For IP (ip_forwarding only) it's ok - 4 hops. > > > For LDP using "mpls_adm" it's ok - 3 hops > > > For LDP using "ldp_linux" IT NEVER WORK!!!!!!!!!! > > >=20 > > > I see that "ldp_linux" only create the LSP for one direction (from= A=20 > > to B)! > > > Like documentation "ldp_linux" create LSPs basead on the L3=20 > > information (so, static routes created above should result in LSPs=20 > > created in two directions)! > > > > > > I follow README.sample commands, but for the sample environment it= =20 > > too don't work! > > > > > > If some one in this mailing list had sucessfull configured=20 > > "ldp_linux" in such environment > > > or like README.sample and the LSPs are dinamically create for two=20 > > directions, please let me know! > > > > > > Morvan. > > > > > > _______________________________________________________________ > > > > > > Don't miss the 2002 Sprint PCS Application Developer's Conference > > > August 25-28 in Las Vegas -=20 > > http://devcon.sprintpcs.com/adp/index.cfm?source=3Dosdntextlink > > > > > > _______________________________________________ > > > mpls-linux-general mailing list > > > mpl...@li... > > > https://lists.sourceforge.net/lists/listinfo/mpls-linux-general > >=20 > > -- > > James R. Leu > >=20 --=20 James R. Leu |
|
From: Markus H. <mh...@as...> - 2003-02-18 11:30:35
|
James R. Leu wrote:
> On Mon, Feb 17, 2003 at 12:55:06AM +0100, Markus Hennig wrote:
>
>>Hi all,
>>
>>any news about this?
>>
>>Looks like with CVS revision 1.16 (local p4 tree @274) its still not
>>working...
>>
>>Any (simple) workaround?
>>
>>Thx in advance,
>>Markus
>
>
> The title has me a bit confused. ldp_linux no longer exists, the only way
> I provide for you to run LDP is with zebra. (using the mplsd daemon).
sorry, i afraid i was confused too...
of course im using the latest cvs version of the ldp patch for zebra and
zebra + mplsd and latest patch for the linux kernel....
>
> As far as I know mplsd can create LSPs correctly for the intended environment
> (LSPs being setup for the LDP Identifier) in DU ordered control. Of course
> not every one understands the ramifications of that statment.
>
> Start out by describing what you are trying to do, what you are observing
> and what you expected to see. Please include a ascii drawing of you setup.
i started with a simple setup, with static routes in zebra.conf and a
simple mplsd.conf:
hostname LSR-2
password pppp
enable password pppp
!
!
mpls ldp
! trace debug
! trace event
! trace normal
! trace notification
! trace path
! trace policy
! trace state
egress connected
address-mode ldp
!
interface lo
!
interface eth1
mpls ldp
!
interface dummy0
!
interface eth2
mpls ldp
!
interface eth3
mpls ldp
!
interface eth0
!
line vty
exec-timeout 0 0
!
there was ldp traffic, but no label assigns to remote routes...
i can't reproduce this behaviour again, because i changed the setup and
use now ospfd too (without static routes) :-)
simple setup with to areas:
area 1: 1 2 7 4
area 2: 3 5 6
now i see the label information and can test them with a ping
(dumping it with a modified uml_switch), but there is a other problem:
mpls_in and mpls_out are constantly growing until around 220 and more
(peak was 682) enries for in and 35 for LER-1 - then they decrease
rapidly to something about 40...
do you have any ideas why?
which debugging output would be usefull (which config)?
btw: the systems runs very unstable (i use a while true loop for mplsd),
especially LER-7 (i set the two egress routes per script and don't use
them in zebra,ospfd,mplsd - otherwise it chrashes immediately)
my testsetup:
.--------LSR_2-----------.
/ \ |
/ LSR_5 |
/ / \ | /
/ / \ | /
----LER_1--------LSR_3 LER_7
\ \ / | \
\ \ / | \
\ LSR_6 |
\ / |
'--------LSR_4-----------'
|----------- MPLS ------------|
thx in advance,
markus
|
|
From: James R. L. <jl...@mi...> - 2003-02-18 15:09:00
|
On Tue, Feb 18, 2003 at 12:30:27PM +0100, Markus Hennig wrote: <snip> > now i see the label information and can test them with a ping > (dumping it with a modified uml_switch), but there is a other problem: > > mpls_in and mpls_out are constantly growing until around 220 and more > (peak was 682) enries for in and 35 for LER-1 - then they decrease > rapidly to something about 40... > > do you have any ideas why? Multipath and merging. > btw: the systems runs very unstable (i use a while true loop for mplsd), > especially LER-7 (i set the two egress routes per script and don't use > them in zebra,ospfd,mplsd - otherwise it chrashes immediately) multipath is bad. I have idea how to support it, but I know for sure it will cause things to behave badly. (as an FYI RFC3036 explicitly says it doesn't support multipath. Any solution that vendors have put in place it proprietary.) > my testsetup: > > .--------LSR_2-----------. > / \ | > / LSR_5 | > / / \ | / > / / \ | / > ----LER_1--------LSR_3 LER_7 > \ \ / | \ > \ \ / | \ > \ LSR_6 | > \ / | > '--------LSR_4-----------' > > |----------- MPLS ------------| Evil. Lets try walking before running. Simplify the topo and let me know if you still see so much instability. Eventually mplsd will be able to support this type of 'complex' topo, but I'm not there yet. Maybe with you help we can nail down some of the real troubled areas. Thanks -- James R. Leu |
|
From: Markus H. <mh...@as...> - 2003-02-18 23:03:05
|
James R. Leu wrote:
> On Tue, Feb 18, 2003 at 12:30:27PM +0100, Markus Hennig wrote:
> <snip>
>
>>now i see the label information and can test them with a ping
>>(dumping it with a modified uml_switch), but there is a other problem:
>>
>>mpls_in and mpls_out are constantly growing until around 220 and more
>>(peak was 682) enries for in and 35 for LER-1 - then they decrease
>>rapidly to something about 40...
>>
>>do you have any ideas why?
>
>
> Multipath and merging.
>
ok
>
>>btw: the systems runs very unstable (i use a while true loop for mplsd),
>>especially LER-7 (i set the two egress routes per script and don't use
>>them in zebra,ospfd,mplsd - otherwise it chrashes immediately)
>
>
> multipath is bad. I have idea how to support it, but I know for sure it
> will cause things to behave badly. (as an FYI RFC3036 explicitly says it
sorry for not knowing this :-)
> doesn't support multipath. Any solution that vendors have put in place it
> proprietary.)
ok, i changed the config for mplsd (ospfd and zebra stays!):
.--------LSR_2 IN:74 OUT:10
/ \
/ LSR_5 IN:119 OUT:10
/ / /
/ IN:32 / /
----LER_1 OUT:5 LSR_3 LER_7 IN:53 OUT:1
IN:77 \ | \
OUT:2 \ | \
IN:98 OUT:5 LSR_6 |
/ |
LSR_4-----------'
IN:17 OUT:7
|----------- MPLS ------------|
>
>>my testsetup:
>>
>> .--------LSR_2-----------.
>> / \ |
>> / LSR_5 |
>> / / \ | /
>> / / \ | /
>>----LER_1--------LSR_3 LER_7
>> \ \ / | \
>> \ \ / | \
>> \ LSR_6 |
>> \ / |
>> '--------LSR_4-----------'
>>
>> |----------- MPLS ------------|
>
>
> Evil. Lets try walking before running. Simplify the topo and let me know
> if you still see so much instability.
>
looks more stable (there are some mystic things with not available nics
during startup of ospf - maybe its ulm_swicth, after a ping they are gone)
numbers of incoming/outgoig see above
all numbers are slow growing (and sometimes decreasing)
> Eventually mplsd will be able to support this type of 'complex' topo, but
its ok, there is no need for
> I'm not there yet. Maybe with you help we can nail down some of the real
> troubled areas.
i do my best...
thx in advance,
markus
>
> Thanks
|
|
From: James R. L. <jl...@mi...> - 2003-02-18 23:41:12
|
See comments within. On Wed, Feb 19, 2003 at 12:02:56AM +0100, Markus Hennig wrote: <snip> > ok, i changed the config for mplsd (ospfd and zebra stays!): I think you will need to change OSPF as well. Without changing OSPF LDP is presented with 2 equal cost OSPF routes, this is the fundamental problem for the multipath case. > > > .--------LSR_2 IN:74 OUT:10 > / \ > / LSR_5 IN:119 OUT:10 > / / / > / IN:32 / / > ----LER_1 OUT:5 LSR_3 LER_7 IN:53 OUT:1 > IN:77 \ | \ > OUT:2 \ | \ > IN:98 OUT:5 LSR_6 | > / | > LSR_4-----------' > IN:17 OUT:7 > > |----------- MPLS ------------| > > > > >>my testsetup: > >> > >> .--------LSR_2-----------. > >> / \ | > >> / LSR_5 | > >> / / \ | / > >> / / \ | / > >>----LER_1--------LSR_3 LER_7 > >> \ \ / | \ > >> \ \ / | \ > >> \ LSR_6 | > >> \ / | > >> '--------LSR_4-----------' > >> > >> |----------- MPLS ------------| > > > > > > Evil. Lets try walking before running. Simplify the topo and let me know > > if you still see so much instability. > > > looks more stable (there are some mystic things with not available nics > during startup of ospf - maybe its ulm_swicth, after a ping they are gone) > > numbers of incoming/outgoig see above > all numbers are slow growing (and sometimes decreasing) > > > Eventually mplsd will be able to support this type of 'complex' topo, but > its ok, there is no need for > > I'm not there yet. Maybe with you help we can nail down some of the real > > troubled areas. > i do my best... > > thx in advance, > markus > > > > > Thanks > > > > > ------------------------------------------------------- > This sf.net email is sponsored by:ThinkGeek > Welcome to geek heaven. > http://thinkgeek.com/sf > _______________________________________________ > mpls-linux-general mailing list > mpl...@li... > https://lists.sourceforge.net/lists/listinfo/mpls-linux-general -- James R. Leu |
|
From: Markus H. <mh...@as...> - 2003-02-19 22:28:14
|
James R. Leu wrote: > See comments within. > > On Wed, Feb 19, 2003 at 12:02:56AM +0100, Markus Hennig wrote: > <snip> > >>ok, i changed the config for mplsd (ospfd and zebra stays!): > > > I think you will need to change OSPF as well. Without changing OSPF > LDP is presented with 2 equal cost OSPF routes, this is the fundamental whats with different costs on the paths, will it work then? whats the easiest way to set different costs ? > problem for the multipath case. > > >> >> .--------LSR_2 IN:74 OUT:10 >> / \ >> / LSR_5 IN:119 OUT:10 >> / / / >> / IN:32 / / >>----LER_1 OUT:5 LSR_3 LER_7 IN:53 OUT:1 >> IN:77 \ | \ >> OUT:2 \ | \ >> IN:98 OUT:5 LSR_6 | >> / | >> LSR_4-----------' >> IN:17 OUT:7 >> >> |----------- MPLS ------------| >> >> |
|
From: James R. L. <jl...@mi...> - 2003-02-19 22:41:18
|
On Wed, Feb 19, 2003 at 10:08:33PM +0100, Markus Hennig wrote: > James R. Leu wrote: > > See comments within. > > > > On Wed, Feb 19, 2003 at 12:02:56AM +0100, Markus Hennig wrote: > > <snip> > > > >>ok, i changed the config for mplsd (ospfd and zebra stays!): > > > > > > I think you will need to change OSPF as well. Without changing OSPF > > LDP is presented with 2 equal cost OSPF routes, this is the fundamental > whats with different costs on the paths, will it work then? > whats the easiest way to set different costs ? You can have OSPF modify it for you: conf t interface eth0 ip ospf cost 15 > > > > problem for the multipath case. > > > > > >> > >> .--------LSR_2 IN:74 OUT:10 > >> / \ > >> / LSR_5 IN:119 OUT:10 > >> / / / > >> / IN:32 / / > >>----LER_1 OUT:5 LSR_3 LER_7 IN:53 OUT:1 > >> IN:77 \ | \ > >> OUT:2 \ | \ > >> IN:98 OUT:5 LSR_6 | > >> / | > >> LSR_4-----------' > >> IN:17 OUT:7 > >> > >> |----------- MPLS ------------| > >> > >> > > > > > ------------------------------------------------------- > This SF.net email is sponsored by: SlickEdit Inc. Develop an edge. > The most comprehensive and flexible code editor you can use. > Code faster. C/C++, C#, Java, HTML, XML, many more. FREE 30-Day Trial. > www.slickedit.com/sourceforge > _______________________________________________ > mpls-linux-general mailing list > mpl...@li... > https://lists.sourceforge.net/lists/listinfo/mpls-linux-general -- James R. Leu |
|
From: Markus H. <mh...@as...> - 2003-02-19 22:59:01
|
James R. Leu wrote: > On Wed, Feb 19, 2003 at 10:08:33PM +0100, Markus Hennig wrote: > >>James R. Leu wrote: >> >>>See comments within. >>> >>>On Wed, Feb 19, 2003 at 12:02:56AM +0100, Markus Hennig wrote: >>><snip> >>> >>>>ok, i changed the config for mplsd (ospfd and zebra stays!): >>> >>> >>>I think you will need to change OSPF as well. Without changing OSPF >>>LDP is presented with 2 equal cost OSPF routes, this is the fundamental >> >>whats with different costs on the paths, will it work then? >>whats the easiest way to set different costs ? > > > You can have OSPF modify it for you: > > conf t > interface eth0 > ip ospf cost 15 ok, i will give it a try and report then in/out numbers here... > > >> >>>problem for the multipath case. >>> >>> >>> >>>> .--------LSR_2 IN:74 OUT:10 >>>> / \ >>>> / LSR_5 IN:119 OUT:10 >>>> / / / >>>> / IN:32 / / >>>>----LER_1 OUT:5 LSR_3 LER_7 IN:53 OUT:1 >>>> IN:77 \ | \ >>>> OUT:2 \ | \ >>>> IN:98 OUT:5 LSR_6 | >>>> / | >>>> LSR_4-----------' >>>> IN:17 OUT:7 >>>> >>>> |----------- MPLS ------------| >>>> >>>> >> >> >> >> >>------------------------------------------------------- >>This SF.net email is sponsored by: SlickEdit Inc. Develop an edge. >>The most comprehensive and flexible code editor you can use. >>Code faster. C/C++, C#, Java, HTML, XML, many more. FREE 30-Day Trial. >>www.slickedit.com/sourceforge >>_______________________________________________ >>mpls-linux-general mailing list >>mpl...@li... >>https://lists.sourceforge.net/lists/listinfo/mpls-linux-general > > |
Thanks for helping keep SourceForge clean.
X