GLBP (3 routers)

[eckzxxiii23]

Hi all,

Please HELP!.. we are planning to rebuild our network. It will consist of 3 Cisco Routers and will run for GLBP.

The R1 and R2 routers will be the primary gateway to run GLBP, the R3  will be dedicated only for a one VLAN (vlan50) but will transfer to any of the R1 or R2 just incase of ISP 3 failure. It is possible for this network proposal? Please help.. what do I need for configuration.

[Vasilii Mikhailovskii]

Hello.

Why do you use GLBP and not HSRP or VRRP? Do you want to load-balance between ISPs?

[eckzxxiii23]

Yes Sir, I want  VLANs 10,20,30 and 40 to share on eitheir ISP1 or ISP2. 

VLAN 50 should be dedicated to only ISP3 while on operation.

[Vasilii Mikhailovskii]

I would recommend to share ISPs per VLAN, like ISP1 for VL10 and 20, ISP2 for VL 30 and 40 (with failover).

In this case you may run HSRP/VRRP making router1(ISP1) active for VL10 and 20, router2 for VL30 and 40; the configuration and behaviour would be predictable and easy to support.

[Seb Rupik]

This is possible using GLBP weighting. Essentially you need R1 and R2 to have their weight reduced below a lower threshhold during normal operation, so that they do not become AVFs.

You will need to add in your existing GLBP config, below are the significant differences that I cooked up in emacs... :)

For config below I assume that each router has a L3 interface in VLAN 50 (gi0/2):
R1: 10.10.50.1
R2: 10.10.50.2
R3: 10.10.50.3

Both R1 and R2 need to track R3's VLAN 50 interface. R3 will shutdown its interface in the event that it loses its connection to its ISP.

Lets assume that R3's ISP interface is gi0/1 and the ISPs P2P IP is 10.100.1.1 :

R1

!
track 1 ip sla 10 reachability
!
ip sla monitor 10
  type echo protocol ipIcmpEcho 10.100.1.1 source-interface gi0/1
!

...or maybe:

!
track 1 interface gi0/1 line-protocol
!

!
interface gi0/2.50
  glbp 50 priority 110
!
event manager applet ethertrack
event track 1 state down
 action 1.0 syslog msg "ISP3 is down; shutting down gi0/2"
 action 2.0 cli command "enable"
 action 3.0 cli command "config t"
 action 4.0 cli command "interface gi0/2"
 action 5.0 cli command "shut"
 action 6.0 cli command "end"
 action 7.0 syslog msg "gi0/2 is now down"
!
event track 1 state up
 action 1.0 syslog msg "ISP3 is up; enabling gi0/2"
 action 2.0 cli command "enable"
 action 3.0 cli command "config t"
 action 4.0 cli command "interface gi0/2"
 action 5.0 cli command "no shut"
 action 6.0 cli command "end"
 action 7.0 syslog msg "gi0/2 is now up"
!

Now for R1 and R2, track the status of R3's Gi0/2 interface:

!
ip sla monitor 10
  type echo protocol ipIcmpEcho 10.10.50.3 source-interface gi0/2
!
track 100 ip sla 10 reachability
!
track 200 list boolean and
  object 100 not
!
interface gi0/2.50
  glbp 50 track 200 decrement 15
  glbp 50 weighting 100 lower 95 upper 105
  glbp 50 priority 100
!

cheers,

Seb.

[eckzxxiii23]

Should R3 be my AVG? and R1 and R2 are AVF?

Thank you very much Seb for your answer, I should really try this... 

And I will post some of my confiigurations...

[Seb Rupik]

Undere the inital state, where ISP3 is functional and R1 and R2 GLBP weight is below threshold, you are correct R3 would be the AVG and R1,R2 would be AVFs, although not forwarding any traffic.

Once ISP3 fails, R2 would assume the AVG (highest IP address) role and both R1 and R2 would become active (forwarding) AVF's. R3 would become disconnected as its Gi0/2 is shutdown.

When ISP3 comes back online, R3 would rejoin the group as an AVF.

cheers,

Seb.

[Vasilii Mikhailovskii]

Ok, what happens when ISP3 recovers after long outage?

How the servers would start using ISP3 router as primary gateway?!

PS: really, don't make your life too complex - use VRRP or HSRP.

[Seb Rupik]

Hi Vasilli,

Once ISP3 comes back online R3 would enable its Gi0/2 interface, rejoining the GLBP group for VLAN50. R1 and R2 would detect R3's G0/2 interface coming back up and lower their GLBP weights below the threshold to ensure that only R3 is forwarding packets for VLAN50.

This is the behaviour that the OP asked for. VLANs 10-40 are to be balanced across ISP1-2 at all times, it's a neat solution.

The VRRP/ HSRP soltuion does not provide this. In the event that VLANs10 & 20 were idle, ISP1 would be paid for and wasted bandwidth.

cheers,

Seb.

[eckzxxiii23]

Thanks Seb and Vasilii for your answer,

Acctually, our first network was on HSRP, when we still have 2 ISP on our network.

VLAN 10 - 40 were only on ISP 1, and VLAN 50 were only dedicated on ISP 2.

But now we are on planning for the ISP 3. My problem on part is, I don't have really experience on GLBP.

I took over this network with the previous Network Ad. without documentations. So if you could post any config specially on the weighting part of  GLBP, I really really appreciate your answer. Thank you very very much.

[Seb Rupik]

The GLBP config makes the following assumptions:
VLAN 50 subnet is 10.10.50.0 /24
VLAN50 G/W address will be 10.10.50.254
The 'inside' router interface is Gi0/2

R1

!
interface Gi0/2.50
  encapsulation dot1Q 50
  ip address 10.10.50.1 255.255.255.0
  glbp 50 preempt
  glbp 50 ip 10.10.50.254
  glbp 50 weighting 100 lower 95 upper 105
!

R2

!
interface Gi0/2.50
  encapsulation dot1Q 50
  ip address 10.10.50.2 255.255.255.0
  glbp 50 preempt
  glbp 50 ip 10.10.50.254
  glbp 50 weighting 100 lower 95 upper 105
!

R3

!
interface Gi0/2.50
  encapsulation dot1Q 50
  ip address 10.10.50.3 255.255.255.0
  glbp 50 preempt
  glbp priority 150
  glbp 50 ip 10.10.50.254
!

You will also need to include the tracking information from above.

I suggest that you lab this up in GNS3 and are happy that it works.

Please rate helpful posts!

cheers,

Seb.