26 mins read
RSVP – soft-preemption/adaptive
1. Intro
Essentially, we are discussing here about the make before break (MBB) feature. You will see MBB is enabled by default for some features, like periodic re-optimization of LSPs. By default, MBB is disabled for LSP preemption, but we can change this with either soft-preemption or adaptive.
Soft preemption attempts to establish a new path for a preempted LSP before tearing down the original LSP. The default behavior is to tear down a preempted LSP first, signal a new path, and then reestablish the LSP over the new path. In the interval between when the path is taken down and the new LSP is established, any traffic attempting to use the LSP is lost. Soft preemption prevents this type of traffic loss. The trade-off is that during the time when an LSP is being soft preempted, two LSPs with their corresponding bandwidth requirements are used until the original path is torn down.
You can configure an LSP to be adaptive when it is attempting to reroute itself. When it is adaptive, the LSP holds onto existing resources until the new path is successfully established and traffic has been cut over to the new LSP. To retain its resources, an adaptive LSP will ensure:
- that the existing path is not torn down prematurely to allows the current traffic to continue flowing while the new path is being set up,
- will avoid double-counting for links that share the new and old paths (using the shared explicit allocation style).
The adaptive feature changes the LSP reservation style. By default, Junos is using the FF (fixed filter) reservation style, which means that if one LSP has two established paths through the network and if both cross the same link, the LSP bandwidth is reserved twice over that link. This might cause an issue if there is not enough logical bandwidth left for the LSP to be established. The adaptive feature changes the reservation to SE (shared explicit). This way the bandwidth will be shared and only accounted once on the common links.
The topology we will use for testing:
Note:
- Based on the addressing being used you will be able to identify the link.
- For example. 17.6.7.7 is a link between router 6 and 7, and this is Router 7.
- Second example. 17.1.2.1 is a link between router 1 and 2, and this is Router 1.
We will start with the following config:
### We will start with one initial Best-Effort LSP:
set protocols mpls label-switched-path R1-to-R7-BEST-EFFORT to 17.0.0.7
set protocols mpls label-switched-path R1-to-R7-BEST-EFFORT bandwidth 600m
set protocols mpls label-switched-path R1-to-R7-BEST-EFFORT priority 5 5
### All interfaces have 1000 Mbps bandwidth:
root@R1# run show rsvp interface
RSVP interface: 2 active
Active Subscr- Static Available Reserved Highwater
Interface State resv iption BW BW BW mark
ge-0/0/1.0 Up 0 100% 1000Mbps 1000Mbps 0bps 600Mbps
ge-0/0/2.0 Up 1 100% 1000Mbps 400Mbps 600Mbps 950Mbps
### Our initial LSP is using 600 Mbps on the main path R1-R3-R6-R7:
root@R1# run show mpls lsp ingress name R1-to-R7-BEST-EFFORT extensive
Ingress LSP: 1 sessions
17.0.0.7
From: 17.0.0.1, State: Up, ActiveRoute: 0, LSPname: R1-to-R7-BEST-EFFORT, LSPid: 19
ActivePath: (primary)
LSPtype: Static Configured, Penultimate hop popping
LoadBalance: Random
Follow destination IGP metric
Encoding type: Packet, Switching type: Packet, GPID: IPv4
LSP Self-ping Status : Enabled
*Primary State: Up
Priorities: 5 5
Bandwidth: 600Mbps
OptimizeTimer: 30
SmartOptimizeTimer: 180
Flap Count: 0
MBB Count: 0
Reoptimization in 5 second(s).
Computed ERO (S [L] denotes strict [loose] hops): (CSPF metric: 3000)
17.1.3.3 S 17.3.6.6 S 17.6.7.7 S
Received RRO (ProtectionFlag 1=Available 2=InUse 4=B/W 8=Node 10=SoftPreempt 20=Node-ID):
17.1.3.3(Label=302000) 17.3.6.6(Label=301920) 17.6.7.7(Label=3)
13 Jul 30 02:28:15.286 CSPF: computation result ignored, new path higher metric
12 Jul 30 02:27:56.448 CSPF: computation result accepted 17.1.3.3 17.3.6.6 17.6.7.7
11 Jul 30 01:27:42.117 CSPF: computation result ignored, new path higher metric
10 Jul 30 01:27:27.540 Selected as active path
9 Jul 30 01:27:27.539 Self-ping ended successfully
8 Jul 30 01:27:26.970 Up
7 Jul 30 01:27:26.970 Self-ping started
6 Jul 30 01:27:26.970 Self-ping enqueued
5 Jul 30 01:27:26.970 Record Route: 17.1.3.3(Label=302000) 17.3.6.6(Label=301920) 17.6.7.7(Label=3)
4 Jul 30 01:27:26.934 LSP-ID: 1 created
3 Jul 30 01:27:26.934 Originate Call
2 Jul 30 01:27:26.934 CSPF: computation result accepted 17.1.3.3 17.3.6.6 17.6.7.7
1 Jul 30 01:26:57.907 CSPF failed: no route toward 17.0.0.7
Created: Thu Jul 30 01:26:57 2026
Total 1 displayed, Up 1, Down 02. By default, no MBB for preemption
We are adding a new LSP R1-to-R7-GOLD, so that it will away R1-to-R7-BEST-EFFORT to an alternate path.
### Let's add one more LSP, same bandwidth but better priority:
root@R1# show | compare
[edit protocols mpls]
label-switched-path R1-to-R7-BEST-EFFORT { ... }
+ label-switched-path R1-to-R6-GOLD {
+ to 17.0.0.6;
+ bandwidth 600m;
+ priority 3 3;
+ }
[edit]
root@R1# commit
commit complete
### There is not enough bandwidth on the links for both or LSPs.
### So, LSP R1-to-R6-GOLD will kick away R1-to-R7-BEST-EFFORT to an alternate path.
root@R1# run show mpls lsp ingress name R1-to-R7-BEST-EFFORT extensive
Ingress LSP: 2 sessions
17.0.0.7
From: 17.0.0.1, State: Up, ActiveRoute: 0, LSPname: R1-to-R7-BEST-EFFORT, LSPid: 19
ActivePath: (primary)
LSPtype: Static Configured, Penultimate hop popping
LoadBalance: Random
Follow destination IGP metric
Encoding type: Packet, Switching type: Packet, GPID: IPv4
LSP Self-ping Status : Enabled
*Primary State: Up
Priorities: 5 5
Bandwidth: 600Mbps
OptimizeTimer: 30
SmartOptimizeTimer: 180
Flap Count: 1
MBB Count: 0
Reoptimization in 0 second(s).
Computed ERO (S [L] denotes strict [loose] hops): (CSPF metric: 5000)
17.1.2.2 S 17.2.4.4 S 17.4.5.5 S 17.5.8.8 S 17.7.8.7 S
Received RRO (ProtectionFlag 1=Available 2=InUse 4=B/W 8=Node 10=SoftPreempt 20=Node-ID):
17.1.2.2(Label=301152) 17.2.4.4(Label=301376) 17.4.5.5(Label=301296) 17.5.8.8(Label=300912) 17.7.8.7(Label=3)
29 Jul 30 02:39:27.594 CSPF failed: no route toward 17.0.0.7
28 Jul 30 02:39:18.862 Selected as active path
27 Jul 30 02:39:18.861 Self-ping ended successfully
26 Jul 30 02:39:17.949 Up
25 Jul 30 02:39:17.949 Self-ping started
24 Jul 30 02:39:17.949 Self-ping enqueued
23 Jul 30 02:39:17.949 Record Route: 17.1.2.2(Label=301152) 17.2.4.4(Label=301376) 17.4.5.5(Label=301296) 17.5.8.8(Label=300912) 17.7.8.7(Label=3)
22 Jul 30 02:39:17.896 LSP-ID: 2 created
21 Jul 30 02:39:17.896 Originate Call
20 Jul 30 02:39:17.896 Clear Call
19 Jul 30 02:39:17.896 CSPF: computation result accepted 17.1.2.2 17.2.4.4 17.4.5.5 17.5.8.8 17.7.8.7
18 Jul 30 02:39:17.896 Deselected as active
17 Jul 30 02:39:17.895 ResvTear received
16 Jul 30 02:39:17.895 17.1.3.1: Down
15 Jul 30 02:39:17.895 17.1.3.3: Requested bandwidth unavailable
14 Jul 30 02:39:17.893 17.1.3.3: Session preempted
[...]
Created: Thu Jul 30 01:26:57 2026
Total 1 displayed, Up 1, Down 0
### A ping was running, and we had a few packets lost:
root@R1> ping 192.168.7.1 routing-instance Cust-A rapid count 100000
[...]
--- 192.168.7.1 ping statistics ---
29084 packets transmitted, 29080 packets received,If we check the LSP logs, we can see what is happening:
- the LSP is preempted
- torn down
- a new path is computed
- the new path is signaled and up.
Let’s rollback the config:
### Delete the R2-to-R8-GOLD LSP:
root@R1# show | compare
[edit protocols mpls]
- label-switched-path R2-to-R8-GOLD {
- to 17.0.0.6;
- bandwidth 600m;
- priority 3 3;
- }
[edit]
root@R1# commit
commit complete
### After we delete the LSP, the initial R1-to-R7-BEST-EFFORT can get back to the original better path.
### As mentioned in the Intro, this re-optimization will trigger MBB:
[edit]
root@R1# run show mpls lsp ingress name R1-to-R7-BEST-EFFORT extensive
Ingress LSP: 1 sessions
17.0.0.7
From: 17.0.0.1, State: Up, ActiveRoute: 0, LSPname: R1-to-R7-BEST-EFFORT, LSPid: 19
ActivePath: (primary)
LSPtype: Static Configured, Penultimate hop popping
LoadBalance: Random
Follow destination IGP metric
Encoding type: Packet, Switching type: Packet, GPID: IPv4
LSP Self-ping Status : Enabled
*Primary State: Up
Priorities: 5 5
Bandwidth: 600Mbps
OptimizeTimer: 30
SmartOptimizeTimer: 180
Flap Count: 1
MBB Count: 1
Reoptimization in 14 second(s).
Computed ERO (S [L] denotes strict [loose] hops): (CSPF metric: 3000)
17.1.3.3 S 17.3.6.6 S 17.6.7.7 S
Received RRO (ProtectionFlag 1=Available 2=InUse 4=B/W 8=Node 10=SoftPreempt 20=Node-ID):
17.1.3.3(Label=302048) 17.3.6.6(Label=301984) 17.6.7.7(Label=3)
44 Jul 30 02:43:22.108 CSPF: computation result ignored, new path higher metric
43 Jul 30 02:43:05.597 CSPF failed: no route toward 17.0.0.7
42 Jul 30 02:43:05.596 Originate make-before-break call: Add skipped CSPF run
41 Jul 30 02:43:05.596 Make-before-break: Cleaned up old instance: Hold dead expiry
40 Jul 30 02:42:53.400 Pending old path instance deletion[2 times, first Jul 30 02:42:24.202]
39 Jul 30 02:41:55.185 Make-before-break: Switched to new instance
38 Jul 30 02:41:55.184 Self-ping ended successfully
37 Jul 30 02:41:54.714 Up
36 Jul 30 02:41:54.714 Self-ping started
35 Jul 30 02:41:54.714 Self-ping enqueued
34 Jul 30 02:41:54.714 Record Route: 17.1.3.3(Label=302048) 17.3.6.6(Label=301984) 17.6.7.7(Label=3)
33 Jul 30 02:41:54.680 LSP-ID: 3 created
32 Jul 30 02:41:54.680 Originate make-before-break call
31 Jul 30 02:41:54.680 CSPF: computation result accepted 17.1.3.3 17.3.6.6 17.6.7.7
30 Jul 30 02:41:54.680 CSPF: Reroute due to re-optimization
29 Jul 30 02:39:27.594 CSPF failed: no route toward 17.0.0.7
28 Jul 30 02:39:18.862 Selected as active path
27 Jul 30 02:39:18.861 Self-ping ended successfully
26 Jul 30 02:39:17.949 Up
25 Jul 30 02:39:17.949 Self-ping started
24 Jul 30 02:39:17.949 Self-ping enqueued
23 Jul 30 02:39:17.949 Record Route: 17.1.2.2(Label=301152) 17.2.4.4(Label=301376) 17.4.5.5(Label=301296) 17.5.8.8(Label=300912) 17.7.8.7(Label=3)
22 Jul 30 02:39:17.896 LSP-ID: 2 created
21 Jul 30 02:39:17.896 Originate Call
20 Jul 30 02:39:17.896 Clear Call
19 Jul 30 02:39:17.896 CSPF: computation result accepted 17.1.2.2 17.2.4.4 17.4.5.5 17.5.8.8 17.7.8.7
18 Jul 30 02:39:17.896 Deselected as active
17 Jul 30 02:39:17.895 ResvTear received
16 Jul 30 02:39:17.895 17.1.3.1: Down
15 Jul 30 02:39:17.895 17.1.3.3: Requested bandwidth unavailable
14 Jul 30 02:39:17.893 17.1.3.3: Session preempted
[...]
Created: Thu Jul 30 01:26:57 2026
Total 1 displayed, Up 1, Down 03. soft-preemption
Configure soft-preemption and test again:
### First, we will add 'soft-preemption' to the Best-Effort LSP:
root@R1# show | compare
[edit protocols mpls label-switched-path R1-to-R7-BEST-EFFORT]
+ soft-preemption;
[edit]
root@R1# commit
commit complete
root@R1# run show rsvp session ingress extensive name R1-to-R7-BEST-EFFORT
Ingress RSVP: 1 sessions
17.0.0.7
From: 17.0.0.1, LSPstate: Up, ActiveRoute: 0
LSPname: R1-to-R7-BEST-EFFORT, LSPpath: Primary
LSPtype: Static Configured
Suggested label received: -, Suggested label sent: -
Recovery label received: -, Recovery label sent: 302208
Resv style: 1 FF, Label in: -, Label out: 302208
Time left: -, Since: Thu Jul 30 04:15:01 2026
Tspec: rate 600Mbps size 600Mbps peak Infbps m 20 M 1500
Port number: sender 2 receiver 60175 protocol 0
Soft preemption desired
Enhanced FRR: Enabled (Downstream)
PATH rcvfrom: localclient
Adspec: sent MTU 1500
Path MTU: received 1500
PATH sentto: 17.1.3.3 (ge-0/0/2.0) 1 pkts
outgoing message state: refreshing, Message ID: 632, Epoch: 10474068
RESV rcvfrom: 17.1.3.3 (ge-0/0/2.0) 1 pkts, Entropy label: Yes
incoming message handle: R-2762/1, Message ID: 2832, Epoch: 10474044
Explct route: 17.1.3.3 17.3.6.6 17.6.7.7
Record route: <self> 17.1.3.3 17.3.6.6 17.6.7.7
Total 1 displayed, Up 1, Down 0
### Let's add the GOLD LSP again and notice the impact:
root@R1# show | compare
[edit protocols mpls]
label-switched-path R1-to-R7-BEST-EFFORT { ... }
+ label-switched-path R1-to-R6-GOLD {
+ to 17.0.0.6;
+ bandwidth 600m;
+ priority 3 3;
+ }
[edit]
root@R1# commit
commit complete
### We can see that this time, MBB was called:
root@R1# run show mpls lsp name R1-to-R7-BEST-EFFORT extensive
Ingress LSP: 2 sessions
17.0.0.7
From: 17.0.0.1, State: Up, ActiveRoute: 0, LSPname: R1-to-R7-BEST-EFFORT, LSPid: 19
ActivePath: (primary)
LSPtype: Static Configured, Penultimate hop popping
LoadBalance: Random
Follow destination IGP metric
Encoding type: Packet, Switching type: Packet, GPID: IPv4
LSP Self-ping Status : Enabled
*Primary State: Up
Priorities: 5 5
Bandwidth: 600Mbps
OptimizeTimer: 30
SmartOptimizeTimer: 180
Flap Count: 4
MBB Count: 6
Reoptimization in 0 second(s).
Computed ERO (S [L] denotes strict [loose] hops): (CSPF metric: 5000)
17.1.2.2 S 17.2.4.4 S 17.4.5.5 S 17.5.8.8 S 17.7.8.7 S
Received RRO (ProtectionFlag 1=Available 2=InUse 4=B/W 8=Node 10=SoftPreempt 20=Node-ID):
17.1.2.2(Label=301216) 17.2.4.4(Label=301440) 17.4.5.5(Label=301360) 17.5.8.8(Label=300976) 17.7.8.7(Label=3)
158 Jul 30 03:45:28.120 CSPF failed: no route toward 17.0.0.7[2 times, first Jul 30 03:45:21.963]
157 Jul 30 03:45:21.963 Originate make-before-break call: Add skipped CSPF run
156 Jul 30 03:45:21.963 Make-before-break: Cleaned up old instance: due to path down
155 Jul 30 03:44:59.348 Pending old path instance deletion
154 Jul 30 03:44:52.109 Make-before-break: Switched to new instance
153 Jul 30 03:44:52.108 Self-ping ended successfully
152 Jul 30 03:44:52.020 Up
151 Jul 30 03:44:52.020 Self-ping started
150 Jul 30 03:44:52.020 Self-ping enqueued
149 Jul 30 03:44:52.020 Record Route: 17.1.2.2(Label=301216) 17.2.4.4(Label=301440) 17.4.5.5(Label=301360) 17.5.8.8(Label=300976) 17.7.8.7(Label=3)
148 Jul 30 03:44:51.951 LSP-ID: 11 created
147 Jul 30 03:44:51.951 Originate make-before-break call
146 Jul 30 03:44:51.951 CSPF: computation result accepted 17.1.2.2 17.2.4.4 17.4.5.5 17.5.8.8 17.7.8.7
145 Jul 30 03:44:51.951 17.1.3.3: Reroute request due to soft preemption received.
[...]
### From our ping we can see there was no packet loss this time:
root@R1> ping 192.168.7.1 routing-instance Cust-A rapid count 100000
PING 192.168.7.1 (192.168.7.1): 56 data bytes
[...]
--- 192.168.7.1 ping statistics ---
16604 packets transmitted, 16604 packets received, 0% packet lossIf we check the LSP logs, we can see what is happening:
- the LSP is soft preempted
- a new path is computed
- MBB is called
- the new path is signaled and up
- switch traffic to new path
- wait 30 seconds (the default), then clean up old instance
4. Adaptive vs preemption
Let’s rollback and do the same using the adaptive feature:
### Replace 'soft-preemption' with 'adaptive':
root@R1# show | compare
[edit protocols mpls label-switched-path R1-to-R7-BEST-EFFORT]
- soft-preemption;
+ adaptive;
[edit]
root@R1# commit
commit complete
### We can see the Shared Explicit allocation style:
root@R1# run show rsvp session ingress
Ingress RSVP: 1 sessions
To From State Rt Style Labelin Labelout LSPname
17.0.0.7 17.0.0.1 Up 0 1 SE - 302224 R1-to-R7-BEST-EFFORT
Total 1 displayed, Up 1, Down 0
root@R1# run show rsvp session ingress extensive
Ingress RSVP: 1 sessions
17.0.0.7
From: 17.0.0.1, LSPstate: Up, ActiveRoute: 0
LSPname: R1-to-R7-BEST-EFFORT, LSPpath: Primary
LSPtype: Static Configured
Suggested label received: -, Suggested label sent: -
Recovery label received: -, Recovery label sent: 302224
Resv style: 1 SE, Label in: -, Label out: 302224
Time left: -, Since: Thu Jul 30 04:21:02 2026
Tspec: rate 600Mbps size 600Mbps peak Infbps m 20 M 1500
Port number: sender 2 receiver 60176 protocol 0
Enhanced FRR: Enabled (Downstream)
PATH rcvfrom: localclient
Adspec: sent MTU 1500
Path MTU: received 1500
PATH sentto: 17.1.3.3 (ge-0/0/2.0) 1 pkts
outgoing message state: refreshing, Message ID: 636, Epoch: 10474068
RESV rcvfrom: 17.1.3.3 (ge-0/0/2.0) 1 pkts, Entropy label: Yes
incoming message handle: R-2764/1, Message ID: 2835, Epoch: 10474044
Explct route: 17.1.3.3 17.3.6.6 17.6.7.7
Record route: <self> 17.1.3.3 17.3.6.6 17.6.7.7
Total 1 displayed, Up 1, Down 0
### Let's add again the GOLD LSP:
root@R1# show | compare
[edit protocols mpls]
label-switched-path R1-to-R7-BEST-EFFORT { ... }
+ label-switched-path R1-to-R6-GOLD {
+ to 17.0.0.6;
+ bandwidth 600m;
+ priority 3 3;
+ }
[edit]
root@R1# commit
commit complete
### We can see that there was no Make-Before-Break:
root@R1# run show mpls lsp name R1-to-R7-BEST-EFFORT extensive
Ingress LSP: 2 sessions
17.0.0.7
From: 17.0.0.1, State: Up, ActiveRoute: 0, LSPname: R1-to-R7-BEST-EFFORT, LSPid: 27
ActivePath: (primary)
LSPtype: Static Configured, Penultimate hop popping
LoadBalance: Random
Follow destination IGP metric
Encoding type: Packet, Switching type: Packet, GPID: IPv4
LSP Self-ping Status : Enabled
*Primary State: Up
Priorities: 5 5
Bandwidth: 600Mbps
OptimizeTimer: 30
SmartOptimizeTimer: 180
Flap Count: 2
MBB Count: 2
Reoptimization in 1 second(s).
Computed ERO (S [L] denotes strict [loose] hops): (CSPF metric: 5000)
17.1.2.2 S 17.2.4.4 S 17.4.5.5 S 17.5.8.8 S 17.7.8.7 S
Received RRO (ProtectionFlag 1=Available 2=InUse 4=B/W 8=Node 10=SoftPreempt 20=Node-ID):
17.1.2.2(Label=301280) 17.2.4.4(Label=301504) 17.4.5.5(Label=301424) 17.5.8.8(Label=301040) 17.7.8.7(Label=3)
68 Jul 30 04:50:58.185 Selected as active path
67 Jul 30 04:50:58.183 Self-ping ended successfully
66 Jul 30 04:50:58.120 Up
65 Jul 30 04:50:58.120 Self-ping started
64 Jul 30 04:50:58.120 Self-ping enqueued
63 Jul 30 04:50:58.119 Record Route: 17.1.2.2(Label=301280) 17.2.4.4(Label=301504) 17.4.5.5(Label=301424) 17.5.8.8(Label=301040) 17.7.8.7(Label=3)
62 Jul 30 04:50:58.021 LSP-ID: 5 created
61 Jul 30 04:50:58.021 Originate Call
60 Jul 30 04:50:58.021 Clear Call
59 Jul 30 04:50:58.021 CSPF: computation result accepted 17.1.2.2 17.2.4.4 17.4.5.5 17.5.8.8 17.7.8.7
58 Jul 30 04:50:58.021 Deselected as active
57 Jul 30 04:50:58.020 ResvTear received
56 Jul 30 04:50:58.020 17.1.3.1: Down
55 Jul 30 04:50:58.020 17.1.3.3: Requested bandwidth unavailable
54 Jul 30 04:50:58.018 17.1.3.3: Session preempted
[...] If adaptive will not do MBB when preemption an LSP, then when to use it? See next section.
5. adaptive correct use
We will leave only the Best Effort LSP configured with no adaptive (we will delete the GOLD LSP).
### Delete GOLD LSP
### Disable link to R2
### Delete adaptive
root@R1# show | compare
[edit interfaces ge-0/0/1]
- disable;
[edit protocols mpls label-switched-path R1-to-R7-BEST-EFFORT]
- adaptive;
[edit protocols mpls]
label-switched-path R1-to-R7-BEST-EFFORT { ... }
- label-switched-path R1-to-R6-GOLD {
- to 17.0.0.6;
- bandwidth 600m;
- priority 3 3;
- }
[edit]
root@R1# commit
commit complete
### We will use the Best-Effort LSP to demonstrate adaptive, notice the path that is used:
root@R1# run show mpls lsp name R1-to-R7-BEST-EFFORT extensive ingress
Ingress LSP: 1 sessions
17.0.0.7
From: 17.0.0.1, State: Up, ActiveRoute: 0, LSPname: R1-to-R7-BEST-EFFORT, LSPid: 33
ActivePath: (primary)
LSPtype: Static Configured, Penultimate hop popping
LoadBalance: Random
Follow destination IGP metric
Encoding type: Packet, Switching type: Packet, GPID: IPv4
LSP Self-ping Status : Enabled
*Primary State: Up
Priorities: 5 5
Bandwidth: 600Mbps
OptimizeTimer: 30
SmartOptimizeTimer: 180
Flap Count: 1
MBB Count: 0
Reoptimization in 25 second(s).
Computed ERO (S [L] denotes strict [loose] hops): (CSPF metric: 12000)
17.1.3.3 S 17.3.6.6 S 17.6.7.7 S
Received RRO (ProtectionFlag 1=Available 2=InUse 4=B/W 8=Node 10=SoftPreempt 20=Node-ID):
17.1.3.3(Label=302448) 17.3.6.6(Label=302224) 17.6.7.7(Label=3)
28 Jul 30 08:10:38.566 Selected as active path
27 Jul 30 08:10:38.564 Self-ping ended successfully
26 Jul 30 08:10:37.894 Up
25 Jul 30 08:10:37.893 Self-ping started
24 Jul 30 08:10:37.893 Self-ping enqueued
23 Jul 30 08:10:37.893 Record Route: 17.1.3.3(Label=302448) 17.3.6.6(Label=302224) 17.6.7.7(Label=3)
22 Jul 30 08:10:37.860 LSP-ID: 2 created
21 Jul 30 08:10:37.860 Originate Call
20 Jul 30 08:10:37.860 CSPF: computation result accepted 17.1.3.3 17.3.6.6 17.6.7.7
19 Jul 30 08:10:23.840 CSPF failed: no route toward 17.0.0.7
18 Jul 30 08:10:23.840 CSPF: link down/deleted: 17.1.2.1(17.0.0.1:357)(17.0.0.1)->17.1.2.2(17.0.0.2:0)(17.0.0.2)
### We will decrease the bandwidth on the R3-R6 link:
root@R3# show | compare
[edit protocols rsvp interface ge-0/0/2.0]
+ bandwidth 500m;
[edit]
root@R3# commit
commit complete
### At this point RSVP should navigate around this link:
root@R3# run show rsvp interface
RSVP interface: 4 active
Active Subscr- Static Available Reserved Highwater
Interface State resv iption BW BW BW mark
ge-0/0/1.0 Up 1 100% 1000Mbps 1000Mbps 0bps 0bps
ge-0/0/2.0 Up 1 100% 500Mbps -100Mbps 600Mbps 950Mbps
ge-0/0/3.0 Up 0 100% 1000Mbps 1000Mbps 0bps 0bps
ge-0/0/4.0 Up 0 100% 1000Mbps 1000Mbps 0bps 600Mbps
### There a valid path to use: R1-R3-R4-R5-R8-R7, but we RSVP can't set it up.
### This is the issue adaptive is solving. We are still using the R3-R6 link:
root@R1# run show mpls lsp name R1-to-R7-BEST-EFFORT extensive ingress
Ingress LSP: 1 sessions
17.0.0.7
From: 17.0.0.1, State: Up, ActiveRoute: 0, LSPname: R1-to-R7-BEST-EFFORT, LSPid: 33
ActivePath: (primary)
LSPtype: Static Configured, Penultimate hop popping
LoadBalance: Random
Follow destination IGP metric
Encoding type: Packet, Switching type: Packet, GPID: IPv4
LSP Self-ping Status : Enabled
*Primary State: Up
Priorities: 5 5
Bandwidth: 600Mbps
OptimizeTimer: 30
SmartOptimizeTimer: 180
Flap Count: 1
MBB Count: 0
Reoptimization in 22 second(s).
Computed ERO (S [L] denotes strict [loose] hops): (CSPF metric: 12000)
17.1.3.3 S 17.3.6.6 S 17.6.7.7 S
Received RRO (ProtectionFlag 1=Available 2=InUse 4=B/W 8=Node 10=SoftPreempt 20=Node-ID):
17.1.3.3(Label=302448) 17.3.6.6(Label=302224) 17.6.7.7(Label=3)
29 Jul 30 08:11:35.913 CSPF failed: no route toward 17.0.0.7[2 times, first Jul 30 08:11:06.767]
28 Jul 30 08:10:38.566 Selected as active path
27 Jul 30 08:10:38.564 Self-ping ended successfully
26 Jul 30 08:10:37.894 Up
25 Jul 30 08:10:37.893 Self-ping started
24 Jul 30 08:10:37.893 Self-ping enqueued
23 Jul 30 08:10:37.893 Record Route: 17.1.3.3(Label=302448) 17.3.6.6(Label=302224) 17.6.7.7(Label=3)
22 Jul 30 08:10:37.860 LSP-ID: 2 created
21 Jul 30 08:10:37.860 Originate Call
20 Jul 30 08:10:37.860 CSPF: computation result accepted 17.1.3.3 17.3.6.6 17.6.7.7
19 Jul 30 08:10:23.840 CSPF failed: no route toward 17.0.0.7
18 Jul 30 08:10:23.840 CSPF: link down/deleted: 17.1.2.1(17.0.0.1:357)(17.0.0.1)->17.1.2.2(17.0.0.2:0)(17.0.0.2)
We have an LSP R1-R3-R6-R7 and we need to set up a new one R1-R3-R4-R5-R8-R7. The problem is the R1-R3 link, our Best Effort LSP it is already using 600 Mpbs on that link. Setting up a new path for the same Best Effort LSP would mean to double allocate 600 Mbps.
You may say that the old and new path is for the same Best Effort LSP, and should not be double counted, but still is. Well, adaptive was created for this exact scenario:
### Configure adaptive:
root@R1# show | compare
[edit protocols mpls label-switched-path R1-to-R7-BEST-EFFORT]
+ adaptive;
[edit]
root@R1# commit
commit complete
### Redo the same test, exactly as we previously did in this section:
### We can see the the old and new LSP can coexist now:
root@R1# run show rsvp session ingress
Ingress RSVP: 2 sessions
To From State Rt Style Labelin Labelout LSPname
17.0.0.7 17.0.0.1 Up 0 1 SE - 302480 R1-to-R7-BEST-EFFORT
17.0.0.7 17.0.0.1 Up 0 1 SE - 302496 R1-to-R7-BEST-EFFORT
Total 2 displayed, Up 2, Down 0
### The bandwidth is no longer double counted:
root@R1# run show rsvp interface
RSVP interface: 2 active
Active Subscr- Static Available Reserved Highwater
Interface State resv iption BW BW BW mark
ge-0/0/1.0 Down 0 100% 1000Mbps 1000Mbps 0bps 600Mbps
ge-0/0/2.0 Up 1 100% 1000Mbps 400Mbps 600Mbps 950Mbps
### The recomputed path is UP and MBB was used:
root@R1# run show mpls lsp name R1-to-R7-BEST-EFFORT extensive ingress
Ingress LSP: 1 sessions
17.0.0.7
From: 17.0.0.1, State: Up, ActiveRoute: 0, LSPname: R1-to-R7-BEST-EFFORT, LSPid: 34
ActivePath: (primary)
LSPtype: Static Configured, Penultimate hop popping
LoadBalance: Random
Follow destination IGP metric
Encoding type: Packet, Switching type: Packet, GPID: IPv4
LSP Self-ping Status : Enabled
*Primary State: Up
Priorities: 5 5
Bandwidth: 600Mbps
OptimizeTimer: 30
SmartOptimizeTimer: 180
Flap Count: 0
MBB Count: 2
Reoptimization in 13 second(s).
Computed ERO (S [L] denotes strict [loose] hops): (CSPF metric: 5000)
17.1.3.3 S 17.3.4.4 S 17.4.5.5 S 17.5.8.8 S 17.7.8.7 S
Received RRO (ProtectionFlag 1=Available 2=InUse 4=B/W 8=Node 10=SoftPreempt 20=Node-ID):
17.1.3.3(Label=302496) 17.3.4.4(Label=301632) 17.4.5.5(Label=301552) 17.5.8.8(Label=301216) 17.7.8.7(Label=3)
39 Jul 30 08:32:08.593 CSPF: computation result ignored, new path no benefit[2 times, first Jul 30 08:31:40.454]
38 Jul 30 08:31:40.454 Originate make-before-break call: Add skipped CSPF run
37 Jul 30 08:31:40.454 Make-before-break: Cleaned up old instance: Hold dead expiry
36 Jul 30 08:31:39.664 Pending old path instance deletion[3 times, first Jul 30 08:30:41.234]
35 Jul 30 08:30:13.092 Make-before-break: Switched to new instance
34 Jul 30 08:30:13.090 Self-ping ended successfully
33 Jul 30 08:30:12.324 Up
32 Jul 30 08:30:12.324 Self-ping started
31 Jul 30 08:30:12.324 Self-ping enqueued
30 Jul 30 08:30:12.324 Record Route: 17.1.3.3(Label=302496) 17.3.4.4(Label=301632) 17.4.5.5(Label=301552) 17.5.8.8(Label=301216) 17.7.8.7(Label=3)
29 Jul 30 08:30:12.271 LSP-ID: 3 created
28 Jul 30 08:30:12.271 Originate make-before-break call
27 Jul 30 08:30:12.271 CSPF: computation result accepted 17.1.3.3 17.3.4.4 17.4.5.5 17.5.8.8 17.7.8.7
26 Jul 30 08:30:12.271 CSPF: Reroute due to re-optimization
### After 30 seconds the old path is deleted:
root@R1# run show rsvp session ingress
Ingress RSVP: 1 sessions
To From State Rt Style Labelin Labelout LSPname
17.0.0.7 17.0.0.1 Up 0 1 SE - 302496 R1-to-R7-BEST-EFFORT
Total 1 displayed, Up 1, Down 0Adaptive may also help in a different scenario, when we have primary/secondary standby paths, it will not double count bandwidth:
### Configure primary/secondary paths:
root@R1# show | compare
[edit protocols mpls label-switched-path R1-to-R7-BEST-EFFORT]
+ primary BLUE-PATH;
+ secondary RED-PATH {
+ standby;
+ }
[edit]
root@R1# commit
commit complete
root@R1# show protocols mpls | display set | match BEST
set protocols mpls label-switched-path R1-to-R7-BEST-EFFORT to 17.0.0.7
set protocols mpls label-switched-path R1-to-R7-BEST-EFFORT bandwidth 600m
set protocols mpls label-switched-path R1-to-R7-BEST-EFFORT priority 5 5
set protocols mpls label-switched-path R1-to-R7-BEST-EFFORT adaptive
set protocols mpls label-switched-path R1-to-R7-BEST-EFFORT primary BLUE-PATH
set protocols mpls label-switched-path R1-to-R7-BEST-EFFORT secondary RED-PATH standby
### Both primary and secondary paths are UP:
root@R1# run show mpls lsp name R1-to-R7-BEST-EFFORT extensive ingress | match state
From: 17.0.0.1, State: Up, ActiveRoute: 0, LSPname: R1-to-R7-BEST-EFFORT, LSPid: 34
*Primary BLUE-PATH State: Up
Standby RED-PATH State: Up
### Configured like this, adaptive will not double count parimary/secondary paths:
root@R1# run show rsvp interface
RSVP interface: 2 active
Active Subscr- Static Available Reserved Highwater
Interface State resv iption BW BW BW mark
ge-0/0/1.0 Down 0 100% 1000Mbps 1000Mbps 0bps 600Mbps
ge-0/0/2.0 Up 1 100% 1000Mbps 400Mbps 600Mbps 950Mbps6. Putting all together
We will keep it short this time, but let’s reconfigure the network so it look like this:
Our Best-Effort LSP will be configured from R2-to-R8. In our test we will add new GOLD LSP that has a higher priority, and this will force the Best-Effort LSP to find a new path. We will study the effects soft-preemption/adaptive has on the Best-Effort LSP:
- With
adaptiveconfigured:
### Notice the Best-Effort LSP has 'adaptive' configured:
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT to 17.0.0.8
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT bandwidth 600m
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT priority 5 5
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT adaptive
### Add the GOLD LSP and notice what happens with the Best-Effort LSP:
Computed ERO (S [L] denotes strict [loose] hops): (CSPF metric: 6000)
17.2.4.4 S 17.4.5.5 S 17.5.6.6 S 17.6.7.7 S 17.7.8.8 S
Received RRO (ProtectionFlag 1=Available 2=InUse 4=B/W 8=Node 10=SoftPreempt 20=Node-ID):
17.2.4.4(Label=302000) 17.4.5.5(Label=301920) 17.5.6.6(Label=302752) 17.6.7.7(Label=299904) 17.7.8.8(Label=3)
40 Jul 31 03:20:36.638 Selected as active path
39 Jul 31 03:20:36.637 Self-ping ended successfully
38 Jul 31 03:20:36.614 Up
37 Jul 31 03:20:36.614 Self-ping started
36 Jul 31 03:20:36.614 Self-ping enqueued
35 Jul 31 03:20:36.614 Record Route: 17.2.4.4(Label=302000) 17.4.5.5(Label=301920) 17.5.6.6(Label=302752) 17.6.7.7(Label=299904) 17.7.8.8(Label=3)
34 Jul 31 03:20:36.559 LSP-ID: 3 created
33 Jul 31 03:20:36.559 Originate Call
32 Jul 31 03:20:36.559 CSPF: computation result accepted 17.2.4.4 17.4.5.5 17.5.6.6 17.6.7.7 17.7.8.8
31 Jul 31 03:20:17.345 CSPF failed: no route toward 17.0.0.8
30 Jul 31 03:20:17.344 Clear Call: CSPF computation failed
29 Jul 31 03:20:17.344 Deselected as active
28 Jul 31 03:20:17.343 ResvTear received
27 Jul 31 03:20:17.343 17.2.4.2: Down
26 Jul 31 03:20:17.343 17.5.6.6: Requested bandwidth unavailable
25 Jul 31 03:20:17.341 17.5.6.6: Session preempted
### We know already that adaptive has no impact on the preemption mechanics and this is what we are seeing.
### Notice there was no soft-preemption / MBB:
- Redo the same test with
soft-preemptionconfigured:
### Remove adaptive and add 'soft-preemption' instead:
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT to 17.0.0.8
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT soft-preemption
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT bandwidth 600m
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT priority 5 5
### Repeat the test:
Computed ERO (S [L] denotes strict [loose] hops): (CSPF metric: 6000)
17.2.4.4 S 17.4.5.5 S 17.5.6.6 S 17.6.7.7 S 17.7.8.8 S
Received RRO (ProtectionFlag 1=Available 2=InUse 4=B/W 8=Node 10=SoftPreempt 20=Node-ID):
17.2.4.4(Label=302048) 17.4.5.5(Label=301968) 17.5.6.6(Label=302816) 17.6.7.7(Label=299920) 17.7.8.8(Label=3)
30 Jul 31 03:30:06.780 CSPF failed: no route toward 17.0.0.8
29 Jul 31 03:29:37.307 Selected as active path
28 Jul 31 03:29:37.306 Self-ping ended successfully
27 Jul 31 03:29:37.142 Up
26 Jul 31 03:29:37.142 Self-ping started
25 Jul 31 03:29:37.142 Self-ping enqueued
24 Jul 31 03:29:37.142 Record Route: 17.2.4.4(Label=302048) 17.4.5.5(Label=301968) 17.5.6.6(Label=302816) 17.6.7.7(Label=299920) 17.7.8.8(Label=3)
23 Jul 31 03:29:37.090 LSP-ID: 2 created
22 Jul 31 03:29:37.090 Originate Call
21 Jul 31 03:29:37.090 CSPF: computation result accepted 17.2.4.4 17.4.5.5 17.5.6.6 17.6.7.7 17.7.8.8
20 Jul 31 03:29:31.370 CSPF failed: no route toward 17.0.0.8
19 Jul 31 03:29:31.369 Clear Call: CSPF computation failed
18 Jul 31 03:29:31.369 Deselected as active
17 Jul 31 03:29:31.368 ResvTear received
16 Jul 31 03:29:31.368 17.2.4.2: Down
15 Jul 31 03:29:31.368 17.5.6.6: Requested bandwidth unavailable
14 Jul 31 03:29:31.368 17.5.6.6: Session preempted
13 Jul 31 03:29:08.320 CSPF failed: no route toward 17.0.0.8[2 times, first Jul 31 03:29:01.352]
12 Jul 31 03:29:01.352 17.5.6.6: Reroute request due to soft preemption received.
### Soft preemption was requested, but not possible because of the double counting on links R2-R4 / R4-R5.
### The LSP was still down for a couple of seconds. - Redo the test with both
soft-preemptionandadaptive:
### Both 'soft-preemption' and 'adaptive' are configured:
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT to 17.0.0.8
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT soft-preemption
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT bandwidth 600m
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT priority 5 5
set protocols mpls label-switched-path R2-to-R8-BEST-EFFORT adaptive
### Now there is no more impact:
Computed ERO (S [L] denotes strict [loose] hops): (CSPF metric: 6000)
17.2.4.4 S 17.4.5.5 S 17.5.6.6 S 17.6.7.7 S 17.7.8.8 S
Received RRO (ProtectionFlag 1=Available 2=InUse 4=B/W 8=Node 10=SoftPreempt 20=Node-ID):
17.2.4.4(Label=302080) 17.4.5.5(Label=302000) 17.5.6.6(Label=302864) 17.6.7.7(Label=299936) 17.7.8.8(Label=3)
25 Jul 31 03:40:05.296 CSPF: computation result ignored, new path no benefit
24 Jul 31 03:40:05.296 Originate make-before-break call: Add skipped CSPF run
23 Jul 31 03:40:05.296 Make-before-break: Cleaned up old instance: due to path down
22 Jul 31 03:39:56.279 Pending old path instance deletion
21 Jul 31 03:39:35.715 Make-before-break: Switched to new instance
20 Jul 31 03:39:35.714 Self-ping ended successfully
19 Jul 31 03:39:35.326 Up
18 Jul 31 03:39:35.326 Self-ping started
17 Jul 31 03:39:35.326 Self-ping enqueued
16 Jul 31 03:39:35.326 Record Route: 17.2.4.4(Label=302080) 17.4.5.5(Label=302000) 17.5.6.6(Label=302864) 17.6.7.7(Label=299936) 17.7.8.8(Label=3)
15 Jul 31 03:39:35.276 LSP-ID: 2 created
14 Jul 31 03:39:35.276 Originate make-before-break call
13 Jul 31 03:39:35.276 CSPF: computation result accepted 17.2.4.4 17.4.5.5 17.5.6.6 17.6.7.7 17.7.8.8
12 Jul 31 03:39:35.276 17.5.6.6: Reroute request due to soft preemption received.
### Adaptive helped by not double counting the bandwidth,
### Soft-preemption called for make-before-break.
### Notice we had a graceful transition from old to new path, with no traffic impact. Conclusion
soft-preemptionwill activate the MBB behavior for preempted LSPs.- Some events like newly configured LSP or link down will trigger LSP Preemption.
- Other event like automatic re-optimization (OptimizeTimer, if enabled) will not cause preemption (to reduce the ripple effect of preemption causing more preemption).
adaptivewill avoid double counting bandwidth, and will switch over traffic using MBB (e.g. during re-optimization).- But, adaptive will not enable MBB for preempted LSPs.
- Adaptive will not double count Primary/Secondary paths.
soft-preemptionandadaptivecan work in conjunction.
External Resources: