Hi,
in our network we've noticed that certain routers are using one of their two upstream links, way more than the other. We've seen peaks of 80-90% utilisation on one link where the other just sits at 15-20%
The topology I am working with is rather simple, looking at the attached image, R1 and R2 are our PE routers, there's l2vpn circuits between the two of them. R3 and R4 are our P routers. For transport we use MPLS and LDP, our IGP is OSPF. Issue is that it seems that only one link is using for forwarding traffic towards R2. It looks that rather than load balancing per stream, the router is load balancing per remote hosts, that is to say that he will use one of the two uplink interfaces to reach certain hosts. and the other to reach certain others.
In this case, From R1 (ASR 907) I am trying to reach R2 (ASR 9906), I can see from R1 that I have two paths towards R2:
show ip route 172.16.128.54
Routing entry for 172.16.128.54/32
Known via "ospf 1", distance 110, metric 111, type intra area
Last update from 172.16.2.36 on TenGigabitEthernet0/4/7, 3d05h ago
Routing Descriptor Blocks:
172.16.2.36, from 172.16.0.223, 3d05h ago, via TenGigabitEthernet0/4/7
Route metric is 111, traffic share count is 1
* 172.16.2.34, from 172.16.0.223, 3d05h ago, via TenGigabitEthernet0/3/7
Route metric is 111, traffic share count is 1
Repair Path: 172.16.2.36, via TenGigabitEthernet0/4/7My first question is: I see that the .36 path is being utilized as a repair path. In this state, will the repair path sit unused or, will it be used to load balance? I couldn't find a clear answer on this.
Show MPLS forwarding-table, further confirms that there's 2 destinations:
show mpls forwarding-table 172.16.128.54 32
Local Outgoing Prefix Bytes Label Outgoing Next Hop
Label Label or Tunnel Id Switched interface
260 64110 172.16.128.54/32 0 Te0/4/7 172.16.2.36
64110 172.16.128.54/32 0 Te0/3/7 172.16.2.34 Another question here is, if the two routers are towards a different upstream router, how can the route be utilizing the same local and outgoing labels for both? Shouldn't those be different?
If I look to CEF however
show ip cef exact-route 172.16.128.24 172.16.128.54
172.16.128.24 -> 172.16.128.54 => label [64110|64110]-(local:260)TAG adj out of TenGigabitEthernet0/3/7, addr 172.16.2.34
It shows that the router is utilising only one link.
This is the configuration of OSPF on R1:
router ospf 1
router-id 172.16.0.84
auto-cost reference-bandwidth 1000000
nsr
nsf ietf
fast-reroute per-prefix enable prefix-priority low
fast-reroute per-prefix remote-lfa area 2 tunnel mpls-ldp
timers throttle spf 150 150 5000
timers throttle lsa 1 1020 5000
passive-interface default
no passive-interface TenGigabitEthernet0/3/7
no passive-interface TenGigabitEthernet0/4/7
no passive-interface Loopback0
no passive-interface Loopback2
no passive-interface Loopback3
maximum-paths 16
bfd all-interfaces
mpls ldp sync
This is the LDP configuration:
mpls label protocol ldp
no mpls ip propagate-ttl
mpls ldp nsr
mpls ldp graceful-restart
mpls ldp session protection
mpls ldp igp sync holddown 2000
mpls ldp discovery hello interval 10
mpls ldp discovery hello holdtime 30
mpls ldp discovery targeted-hello interval 15
mpls ldp discovery targeted-hello holdtime 120
mpls ldp discovery targeted-hello accept
multilink bundle-name authenticated
l2vpn
Reading online I cannot get a definite answer on how load balacing works in this scenario, some says that cisco routers by default do load balancing on MPLS, others that in this scenario load balancing can only be achieved using fat/entropy labels.
Could anyone help?
Thanks
