EAPS primary with ERPS subring

  • 0
  • 1
  • Question
  • Updated 4 weeks ago
  • In Progress
Hello!  We have EAPS ring with 4 Extreme switches: E1-E2-E3-E4-E1. E1 is the master.
2 Huawei ERPS switches have to be connected to the original EAPS ring, forming additional ERPS ring: E2-H1-H2-E3. Should we run both ERPS and EAPS on E2 and E3? If so could you please give me (or point me to) an example of configuration (perhaps with detailed EXOS commands) of such a scenario (especially for E2 and E3 mixed ERPS and EAPS sections)? What about CFM? Is it necessary in this case? May be someone could share a GNS3 project with this concept? Thanks!
Photo of Dawid


  • 70 Points

Posted 2 months ago

  • 0
  • 1
Photo of Eric Burke

Eric Burke

  • 1,794 Points 1k badge 2x thumb
Dawid, based on a prior response to a similar question, it looks like ERPS may be the better/more holistic solution.  https://community.extremenetworks.com/extreme/topics/eaps_or_erps_differences
Photo of EtherMAN

EtherMAN, Embassador

  • 6,186 Points 5k badge 2x thumb
here is a sticky .. you cant run EAPS and ERPS on same port.  So hopefully you have additional ports you can use on e2 and e3.  I will be watching you progress with this as we are in our lab now trying to get ERPS working between two Extreme 430's and a Ciena 3930.  We are not having any luck so far.  I dont want to hijack your post so good luck.  I am going to start another thread soon if we run out of options and can't figure out how to make these guys play nice together. 
Photo of EtherMAN

EtherMAN, Embassador

  • 6,186 Points 5k badge 2x thumb
OK we have completed our lab and have some insight to share... Our setup was a pair of 430 connected to each other and then a 3930 connected to each of them making a 3 point ring. 

For weeks we tried many different ways to make this work including using CFM as possibly a common way to get them to work together.  After many frustrating hours and posting this to the Ciena forums an individual shared the key to making this work.  

g.8032 standard uses the destination mac in the R-APS messages exchanged between common ring nodes.  From a sniff trace mirroring the control vlan we can see the Ciena was using

Frame 140: 60 bytes on wire (480 bits), 60 bytes captured (480 bits) on interface 0
Ethernet II, Src: Ciena_80:30:ea (9c:7a:03:80:30:ea), Dst: Itu-T_00:00:65 (01:19:a7:00:00:65)
    Destination: Itu-T_00:00:65 (01:19:a7:00:00:65)
    Source: Ciena_80:30:ea (9c:7a:03:80:30:ea)

They allow you to set a ring id using a numerical characters 1-255... this includes using zero's in front of the number you choose.  Extreme to my knowledge does not allow you to change this ... The key is the last field in the destination mac address... so as you can see it is 65 with the Ciena being the ERPS owner.

here is what our 430's were using 

Frame 142: 60 bytes on wire (480 bits), 60 bytes captured (480 bits) on interface 0
Ethernet II, Src: ExtremeN_8b:e2:05 (00:04:96:8b:e2:05), Dst: Itu-T_00:00:01 (01:19:a7:00:00:01)
    Destination: Itu-T_00:00:01 (01:19:a7:00:00:01)
        Address: Itu-T_00:00:01 (01:19:a7:00:00:01)
         Source: ExtremeN_8b:e2:05 (00:04:96:8b:e2:05)
    Type: IEEE 802.1ag Connectivity Fault Management (CFM) protocol (0x8902)
CFM EOAM 802.1ag/ITU Protocol, Type Ring-Automatic Protection Switching (R-APS)

Since you cant change the Extreme side and we can change the Ciena we first use 1... which did not work and then used 01 which did work..  We now have a G.8032 ring where the Ciena is the owner and the Extreme is transit.  This needs to be changed on the Extreme side in my option.  First how does Exteme choose the last field in the mac address?  If we did a second ring using same switches would the next one be 02?  

Ciena did not have this documented anywhere and if it were not for a senior support engineer who had worked this before jumping in we would be forced to trying to pass an EAPS ring through our Ciena gear and not having optimal fail-over times.  

I have sniff traces and more detail i can share if Extreme TAC or Engineering want to engage.