Mailing List Archive: Routes being ignored with MLX4

Routes being ignored with MLX4

Oct 1, 2014, 12:19 PM

Post #1 of 6 (2453 views)

Hi,

We are new to Brocade and we are experiencing an unexpected behavihor:

We have an MLX 4 with 5.6b, receiving a full routing table, including
default route from our ISP.

We are receiving multiple routes from our internal devices like
192.168.0.0/25, 192.168.0.128/25 and aggregating with aggregate-address
192.168.0.0/24 summary-only.

We have this same setup with other networks that work perfectly, however
before had this same issue and worked intermitently. This one give us
always TTL exceeded with ping test from outside and in a traceroute we can
see it looping towards our ISP. From the router we are able to ping the
IPs.

We tried to advertise the /24 route from another device with iBGP, however
it's ignored and the traffic is still is routed to our ISP.

Even after configuring an static route to an internal device, it's ignored
as well and the 0.0.0.0/0 route is used to route traffic.

If we poing the default route to null0 we lose the TTL exceeded messages
but we lose connectivity as well. We pointed a static to the device where
our /25s and /24 is being advertised right now (192.168.0.26) and
everything started working...

We thought about no free CAM space to install routes, however we don't have
any warning/error message in logs and we have 50 free routes.

May this this be related with classless / classful routing implementations?
We find very weird to have a more specific route ignored to a
default-route, even when a "show ip route" shows the more specific one
choosen. We never had this behavior before.

Thank you in advance for your help!

--
Best Regards,
JosÃ©

Re: Routes being ignored with MLX4 [ In reply to ]

mkallen at gmail

Oct 1, 2014, 1:32 PM

Post #2 of 6 (2427 views)

Permalink

Jose', first thing to check are ip cache and ip route system max commands.
Default settings for those are not enough for a full table.

Mike
On Oct 1, 2014 12:22 PM, "JosÃ© Santos" <santosmcjose@gmail.com> wrote:

> Hi,
>
> We are new to Brocade and we are experiencing an unexpected behavihor:
>
> We have an MLX 4 with 5.6b, receiving a full routing table, including
> default route from our ISP.
>
> We are receiving multiple routes from our internal devices like
> 192.168.0.0/25, 192.168.0.128/25 and aggregating with aggregate-address
> 192.168.0.0/24 summary-only.
>
> We have this same setup with other networks that work perfectly, however
> before had this same issue and worked intermitently. This one give us
> always TTL exceeded with ping test from outside and in a traceroute we can
> see it looping towards our ISP. From the router we are able to ping the
> IPs.
>
> We tried to advertise the /24 route from another device with iBGP, however
> it's ignored and the traffic is still is routed to our ISP.
>
> Even after configuring an static route to an internal device, it's ignored
> as well and the 0.0.0.0/0 route is used to route traffic.
>
> If we poing the default route to null0 we lose the TTL exceeded messages
> but we lose connectivity as well. We pointed a static to the device where
> our /25s and /24 is being advertised right now (192.168.0.26) and
> everything started working...
>
> We thought about no free CAM space to install routes, however we don't
> have any warning/error message in logs and we have 50 free routes.
>
> May this this be related with classless / classful routing
> implementations? We find very weird to have a more specific route ignored
> to a default-route, even when a "show ip route" shows the more specific one
> choosen. We never had this behavior before.
>
> Thank you in advance for your help!
>
> --
> Best Regards,
> JosÃ©
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>

Re: Routes being ignored with MLX4 [ In reply to ]

santosmcjose at gmail

Oct 1, 2014, 7:28 PM

Post #3 of 6 (2427 views)

Permalink

Mike and David,

Thanks for your input.

MLX series support ~512K these are our values:

cam-partition profile ipv4-ipv6-2
system-max ip-cache 524288
system-max ip-route 524288

We don't have any warning/error about not having free resources to install
routes yet, we have about 50-100 free routes. We have planned to have our
MLX replaced with a XMR soon however we would like to have a fix for our
issue before that since the routes seem to be correctly installed in the
routing table but due to the loop (which originates TTL expired), default
route seems to be used instead.

Thanks!
JosÃ©

On 1 October 2014 21:32, Mike Allen <mkallen@gmail.com> wrote:

> Jose', first thing to check are ip cache and ip route system max
> commands. Default settings for those are not enough for a full table.
>
> Mike
> On Oct 1, 2014 12:22 PM, "JosÃ© Santos" <santosmcjose@gmail.com> wrote:
>
>> Hi,
>>
>> We are new to Brocade and we are experiencing an unexpected behavihor:
>>
>> We have an MLX 4 with 5.6b, receiving a full routing table, including
>> default route from our ISP.
>>
>> We are receiving multiple routes from our internal devices like
>> 192.168.0.0/25, 192.168.0.128/25 and aggregating with aggregate-address
>> 192.168.0.0/24 summary-only.
>>
>> We have this same setup with other networks that work perfectly, however
>> before had this same issue and worked intermitently. This one give us
>> always TTL exceeded with ping test from outside and in a traceroute we can
>> see it looping towards our ISP. From the router we are able to ping the
>> IPs.
>>
>> We tried to advertise the /24 route from another device with iBGP,
>> however it's ignored and the traffic is still is routed to our ISP.
>>
>> Even after configuring an static route to an internal device, it's
>> ignored as well and the 0.0.0.0/0 route is used to route traffic.
>>
>> If we poing the default route to null0 we lose the TTL exceeded messages
>> but we lose connectivity as well. We pointed a static to the device where
>> our /25s and /24 is being advertised right now (192.168.0.26) and
>> everything started working...
>>
>> We thought about no free CAM space to install routes, however we don't
>> have any warning/error message in logs and we have 50 free routes.
>>
>> May this this be related with classless / classful routing
>> implementations? We find very weird to have a more specific route ignored
>> to a default-route, even when a "show ip route" shows the more specific one
>> choosen. We never had this behavior before.
>>
>> Thank you in advance for your help!
>>
>> --
>> Best Regards,
>> JosÃ©
>>
>> _______________________________________________
>> foundry-nsp mailing list
>> foundry-nsp@puck.nether.net
>> http://puck.nether.net/mailman/listinfo/foundry-nsp
>>
>

Re: Routes being ignored with MLX4 [ In reply to ]

martijnschmidt at i3d

Oct 2, 2014, 2:56 AM

Post #4 of 6 (2413 views)

Permalink

Hi Jose,

According to page 3133 of the NetIron manual, the MLX platform doesn't
support 512k IPv4 routes unless you are using the "ipv4"
cam-partitioning profile which shuts down IPv6 and VRF completely. If
you're using "ipv4-ipv6-2" you would have a maximum of 480k IPv4 routes,
8k IPv6 routes, and no VRF support.

Now I don't know about your routing table, but mine's sitting well over
500k routes.

Best regards,

Martijn Schmidt
i3D.net NOC

On 10/02/2014 04:28 AM, José Santos wrote:
>
> Mike and David,
>
> Thanks for your input.
>
> MLX series support ~512K these are our values:
>
> cam-partition profile ipv4-ipv6-2
> system-max ip-cache 524288
> system-max ip-route 524288
>
> We don't have any warning/error about not having free resources to
> install routes yet, we have about 50-100 free routes. We have planned
> to have our MLX replaced with a XMR soon however we would like to have
> a fix for our issue before that since the routes seem to be correctly
> installed in the routing table but due to the loop (which originates
> TTL expired), default route seems to be used instead.
>
> Thanks!
> José
>
>
> On 1 October 2014 21:32, Mike Allen <mkallen@gmail.com
> <mailto:mkallen@gmail.com>> wrote:
>
> Jose', first thing to check are ip cache and ip route system max
> commands. Default settings for those are not enough for a full table.
>
> Mike
>
> On Oct 1, 2014 12:22 PM, "José Santos" <santosmcjose@gmail.com
> <mailto:santosmcjose@gmail.com>> wrote:
>
> Hi,
>
> We are new to Brocade and we are experiencing an unexpected
> behavihor:
>
> We have an MLX 4 with 5.6b, receiving a full routing table,
> including default route from our ISP.
>
> We are receiving multiple routes from our internal devices
> like 192.168.0.0/25 <http://192.168.0.0/25>, 192.168.0.128/25
> <http://192.168.0.128/25> and aggregating with
> aggregate-address 192.168.0.0/24 <http://192.168.0.0/24>
> summary-only.
>
> We have this same setup with other networks that work
> perfectly, however before had this same issue and worked
> intermitently. This one give us always TTL exceeded with ping
> test from outside and in a traceroute we can see it looping
> towards our ISP. From the router we are able to ping the IPs.
>
> We tried to advertise the /24 route from another device with
> iBGP, however it's ignored and the traffic is still is routed
> to our ISP.
>
> Even after configuring an static route to an internal device,
> it's ignored as well and the 0.0.0.0/0 <http://0.0.0.0/0>
> route is used to route traffic.
>
> If we poing the default route to null0 we lose the TTL
> exceeded messages but we lose connectivity as well. We pointed
> a static to the device where our /25s and /24 is being
> advertised right now (192.168.0.26) and everything started
> working...
>
> We thought about no free CAM space to install routes, however
> we don't have any warning/error message in logs and we have 50
> free routes.
>
> May this this be related with classless / classful routing
> implementations? We find very weird to have a more specific
> route ignored to a default-route, even when a "show ip route"
> shows the more specific one choosen. We never had this
> behavior before.
>
> Thank you in advance for your help!
>
> --
> Best Regards,
> José
>
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net <mailto:foundry-nsp@puck.nether.net>
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>
>
>
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp

Best regards,

Martijn Schmidt
http://www.i3D.net

i3D.net is a private company registered in The Netherlands at Meent 93b, Rotterdam. Registration #: 14074337 - VAT # NL 8202.63.886.B01. i3D.net is CDSA certified on Content Protection and Security and provides hosting from 16 global ISO-certified datacenters. We are ranked in the Deloitte Technology Fast 500 EMEA as one of the fastest growing technology companies.

Re: Routes being ignored with MLX4 [ In reply to ]

lists at openunix

Oct 2, 2014, 8:46 AM

Post #5 of 6 (2401 views)

Permalink

Hello,

this could be bogon filtering.

You can try to disable it and afterwards reinstall thge route manually:

no ip martian filtering-on

Best regards,

Franz Georg KÃ¶hler

Am 01.10.14 um 21:19 schrieb JosÃ© Santos:
> Hi,
>
> We are new to Brocade and we are experiencing an unexpected behavihor:
>
> We have an MLX 4 with 5.6b, receiving a full routing table, including
> default route from our ISP.
>
> We are receiving multiple routes from our internal devices like
> 192.168.0.0/25, 192.168.0.128/25 and aggregating with aggregate-address
> 192.168.0.0/24 summary-only.
>
> We have this same setup with other networks that work perfectly, however
> before had this same issue and worked intermitently. This one give us
> always TTL exceeded with ping test from outside and in a traceroute we can
> see it looping towards our ISP. From the router we are able to ping the
> IPs.
>
> We tried to advertise the /24 route from another device with iBGP, however
> it's ignored and the traffic is still is routed to our ISP.
>
> Even after configuring an static route to an internal device, it's ignored
> as well and the 0.0.0.0/0 route is used to route traffic.
>
> If we poing the default route to null0 we lose the TTL exceeded messages
> but we lose connectivity as well. We pointed a static to the device where
> our /25s and /24 is being advertised right now (192.168.0.26) and
> everything started working...
>
> We thought about no free CAM space to install routes, however we don't have
> any warning/error message in logs and we have 50 free routes.
>
> May this this be related with classless / classful routing implementations?
> We find very weird to have a more specific route ignored to a
> default-route, even when a "show ip route" shows the more specific one
> choosen. We never had this behavior before.
>
> Thank you in advance for your help!
>

_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

Re: Routes being ignored with MLX4 [ In reply to ]

santosmcjose at gmail

Oct 3, 2014, 7:34 AM

Post #6 of 6 (2418 views)

Permalink

Hi,

Thank you all for your help. Even getting warnings about only 50 to 100
routes in router logs, we tought that wouldn't cause this behavior since
technically we still had free routes.

We ended up filterning /24 routes and adding a default route to reduce the
full routing table and everything is fine now. This way we will be able to
use this equipment before being able to migrate to a 1M routes equipment.

If any of you in the future has this issue, I recommend the following:
Issue an "show ip route ... detail" and you'll be able to see the route,
however it will not be installed in CAM partition if in field "Cam:Index"
you see "INVALID CAM".

Regards,
JosÃ©

On 2 October 2014 16:46, Franz Georg KÃ¶hler <lists@openunix.de> wrote:

> Hello,
>
>
> this could be bogon filtering.
>
> You can try to disable it and afterwards reinstall thge route manually:
>
> no ip martian filtering-on
>
>
> Best regards,
>
> Franz Georg KÃ¶hler
>
> Am 01.10.14 um 21:19 schrieb JosÃ© Santos:
> > Hi,
> >
> > We are new to Brocade and we are experiencing an unexpected behavihor:
> >
> > We have an MLX 4 with 5.6b, receiving a full routing table, including
> > default route from our ISP.
> >
> > We are receiving multiple routes from our internal devices like
> > 192.168.0.0/25, 192.168.0.128/25 and aggregating with aggregate-address
> > 192.168.0.0/24 summary-only.
> >
> > We have this same setup with other networks that work perfectly, however
> > before had this same issue and worked intermitently. This one give us
> > always TTL exceeded with ping test from outside and in a traceroute we
> can
> > see it looping towards our ISP. From the router we are able to ping the
> > IPs.
> >
> > We tried to advertise the /24 route from another device with iBGP,
> however
> > it's ignored and the traffic is still is routed to our ISP.
> >
> > Even after configuring an static route to an internal device, it's
> ignored
> > as well and the 0.0.0.0/0 route is used to route traffic.
> >
> > If we poing the default route to null0 we lose the TTL exceeded messages
> > but we lose connectivity as well. We pointed a static to the device where
> > our /25s and /24 is being advertised right now (192.168.0.26) and
> > everything started working...
> >
> > We thought about no free CAM space to install routes, however we don't
> have
> > any warning/error message in logs and we have 50 free routes.
> >
> > May this this be related with classless / classful routing
> implementations?
> > We find very weird to have a more specific route ignored to a
> > default-route, even when a "show ip route" shows the more specific one
> > choosen. We never had this behavior before.
> >
> > Thank you in advance for your help!
> >
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp
>

Mailing List Archive

Mailing List Archive

Attached Files: