SMLT/MLT/VLACP and routing between 8600 and 5510 stack

[Dorian]

Hello everyone. I recently fall on a strange bug concerning the routing between an ERS 5510 stack of 3 switch and my core (two passport 8600 in IST).

Concerning the configuration I have:
SMLT, SLPP, VRRP and VLACP on the 8600 side.
MLT, VLACP and single IP on the ERS5510 side.

Concerning the IP addressing I have something like that:
10.128.90.40/29
10.128.90.41 for the ERS
10.128.90.42 for Passport 1
10.128.90.43 for Passport 2
10.128.90.44 for Passport 3

When I make a show sys topology, everything is fine. But I can't ping other equipements (from Passport to ERS, or from ERS to VRRP or physical address of the Passport).

When I change the IP range to something like: 10.128.100.0/24
Everything seems fine, I can ping all addresses.

It is a known limitation ?
There is some hidden problem somewhere ?

Concerning the architecture choice, this link is the link between our 2 datacenter. The historical one is quite big (Passport 8600), the new one is very small (20 servers), that's why we are only using 5510.

Regards,

[Dominik]

Hi Dorian,

wich SW version do you run on your ERS8600 and ERS5510 ?
I never personal had any trouble with VRRP on /29 networks.
You can try to use RSMLT instead of VRRP. RSMLT gives you the same functionality as VRRP and it is
easy to set up and you can save one IP address on the subnet.. Maybe you have no prblems with RSMLT on the 10.128.90.0/29 network.

Good Luck

[Dorian]

My  Passport version is totally outdated: 4.1.7.2, I will change it before end of march. My ERS is running a 6.2.3.011.

I will make some tests with RSMLT.

[Michael McNamara]

There could be any number of explanations... in theory it should work fine.

I would suspect you have a configuration issue somewhere.

Good Luck!

[Dorian]

I made more tests and everything works now. I made two changes:

- I disabled SLPP on the Edge switch. That was a configuration issue. I have to kick some ass.
- I have to transform my  netmask to have a /24. For me it is not logical. But it works (and doesn't work with a /29 netmask).

I read the two posts of your blog, and my configuration is quite the same. I opened a case to my support for that.

After some investigations I have always a problem somewhere. My aggregation link is not working as expected. Each link works individually, but when I connect the two links, I loose around 50% of my trafic (so  for me only one link is working).

When I check topology, everything is fine, so for me the physical layer seems to be ok.

[Dorian]

Ok I have made more tests.

First this is my configuration of my 5510:

Code: [Select]

vlan create 924
vlan name 924 "INTERCO-BST-VNX"
vlan ports 1/2 tagging tagAll filter-unregistered-frames disable
vlan ports 2/2 tagging tagAll filter-unregistered-frames disable
vlan members 924 1/2,2/2

mlt 1 name "INTERCO_BST_VNX" enable member 1/2,2/2

interface vlan 924
ip address 10.128.100.6 255.255.255.0 3

vlacp enable
vlacp hold_time 0
interface FastEthernet ALL
vlacp port 1/2,2/2,3/2 timeout short
vlacp port 1/2,2/2,3/2 timeout-scale 5
vlacp port 1/2,2/2 enable
Passport configuration:

Code: [Select]

vlan 924 create byport 1 name "INTERCO-VBST-VNX"
vlan 924 add-mlt 1
vlan 924 ports add 1/1,2/1,7/2 member portmember
vlan 924 ip create 10.128.100.1/255.255.255.0 mac_offset 21
vlan 924 ip vrrp 124 address 10.128.100.3
vlan 924 ip vrrp 124 priority 200

ethernet 7/2 default-vlan-id 924
ethernet 7/2 slpp packet-rx enable
ethernet 7/2 stg 1 stp disable
ethernet 7/2 smlt 100 create
ethernet 7/2 vlacp  enable
ethernet 7/2 vlacp  fast-periodic-time 500
ethernet 7/2 vlacp  timeout short
ethernet 7/2 vlacp  timeout-scale 5

slpp add 924
I have quite the same configuration on my second passport.

When I enable the link between my ERS and one Passport, it works fine.
When I enable two links, I have some strange behavior:
- "show sys topology"  is ok.
- ping from passport to ERS are OK.
- some hosts connected to my ERS are not reachable from system connected to the Passport.

=> It seems that some trafic is dropped/blocked somewhere. If somenone has any idea about a troubleshooting strategy, he is welcome.

(I try to improve my English, but it's not very good Sorry for that)

 

[Michael McNamara]

Hi Dorian,

You can refer to this post for more information about how to configure the ERS 8600s for SMLT;
http://blog.michaelfmcnamara.com/2011/12/avaya-split-multilink-trunking-smlt-layer-2-trunking/

In looking at the information you've provided it doesn't appear that you've configured the ERS 8600 ports as trunk (802.1q) ports. Unfortunately that's probably not your only issue... you should issue the following commands;


config ethernet 7/2 perform-tagging enable


If you'd like additional help you'll need to provide more information. I would need to see the following output;


show mlt ist info
show smlt info
show port info vlacp port 7/2
show log file tail


You can find additional troubleshooting steps in the post I've linked to above.

Cheers!

[raglann]

hi,

I see you are using hardening like VLACP and SLPP, but you havent configured any values to them.

VLACP: Configure the same ethertype on both sides (8600 to 5500) at this moment the configuration is default. With a default ethertype on all interfaces vlacp has no use.
SLPP: You should configure a slpp rx-threshold which is different on both Passports for SLPP to act predictable. So for example rx-threshold 5 on PP1 and rx-threshold 50 on PP2. By doing this you will always have one uplink up and running in a faulty situation.

Do you have any hardening configured on the IST interfaces?

I strongly reccomend upgrading since v4.1 is well known to IST bugs.

Regards,

Rag

[Dorian]

Hello,

some answers and some questions.

Concerning SLPP, We have a configuration with 5 and 30 (Ok it's not 50, but I think 30 will do the job). Anyway, I think I will disable it, My ERS 5510 is not really an edge switch, but a core one. Is behavior is quite symetric with my passport one.

Concerning VLACP: We have the same ether type: 0x8103 on both side (Passport 8600/ ERS 5510). However the ethernet mac address is different: 01:80:c2:00:11:00 for the passport and 00:00:00:00:00:00 for the ERS.

I have a "piste" (don't find the english word for that, track does't seem good). My provider sent me an email saying that their equipement is dropping some LACP traffic (I think he saw MLT or VLACP and thought it was LACP).

What is exactly the purpose of the ethernet mac address defined in VLACP ?

[Michael McNamara]

The default VLACP address is 01:80:c2:00:11:00 but Avaya recommends that you use 01:80:c2:00:00:0f.

You just have to configure the correct address on both sides of the link.

Good Luck!

[Dorian]

If I understand correctly this document:

http://support.avaya.com/css/P8/documents/100101601

VLACP is used to detect end-to-end link failures on direct point-to-point interfaces. This
is accomblished by each switch transmitting VLACP PDU’s at a set timer interval in
order for a llink to maintain a ‘link-up’ state. For all direct connected point-to-point links,
use the reserved multicast MAC address of 01:80:c2:00:00:0f. For end-to-end
connnections traversing intermediate networks, use the default VLACP MAC address
01:80:c2:00:11:00.

I'm in the "end-to-end connection" case => I have to use 01:80:c2:00:11:00

[Dorian]

I received my answer, all multicast traffic (including VLACP) is blocked by my ISP...
=> VLACP kicked in, cut the link, bybye connectivity.

I will try to convince them to allow some mac address (and made the appropriate change for vlacp...)