The Packet Path With Partha

15/10/2026

TI-LFA FAST REROUTE: A PRACTICAL VALIDATION CHECKLIST

A backup path in the configuration is not enough. Prove protection with controlled failure testing:

✅ Identify the protected primary next hop
✅ Verify the loop-free repair path
✅ Inspect repair SIDs and the imposed label stack
✅ Test link, node and SRLG failures
✅ Measure packet loss and convergence time
✅ Check ECMP and microloop behavior
✅ Validate L2VPN, L3VPN and SR Policy services
✅ Capture forwarding evidence before and after failure

The goal is simple: predictable traffic forwarding when the network changes.

Which TI-LFA test do you always include in your lab?

14/10/2026

Day 9 of the 22-day SR-MPLS Learning Journey is live 🎥

So far every Prefix-SID has been configured the obvious way: log in to the router, find the loopback, type an index. Today: the two cases where that stops working.

1. Scale. On ten routers it's twenty minutes and a spreadsheet. On a thousand, every index is a change window, and one typo is a duplicate SID that corrupts forwarding.
2. A router that CANNOT advertise a Prefix-SID at all — an old platform, or a whole region still on LDP that nobody may touch this quarter.

Both have the same answer: the SR Mapping Server. One router advertises SIDs on behalf of the others.

The lab is a real five-router IOS XR capture, and each router has a different SRGB on purpose. That looks like a mistake. It's the proof that the Mapping Server advertises an index, never a label.

One line of configuration, on one router:
1.1.1.1/32 index 1 range 100
A hundred Prefix-SIDs, on routers where nobody typed anything.

The traceroute shows it: one destination, four different labels on the way, one unchanging index. The index is the contract. The label is each router's private business.

16 minutes, full lesson 👇
https://youtu.be/9x-Cf7Zi5GU

🔔 Subscribe on YouTube for a new lesson every day:
https://www.youtube.com/?sub_confirmation=1

👍 Like this page for daily lesson summaries:
https://www.facebook.com/ThePacketPathWithPartha

📺 Full 22-day playlist:
https://www.youtube.com/playlist?list=PLYQHPRJxy5xk

13/10/2026

An SR Policy may be “up” while the wrong candidate path is active.

Candidate-path selection checklist:
✅ Confirm the exact color + endpoint tuple
✅ List all local, BGP and PCEP candidate paths
✅ Compare preference values
✅ Check validity and disqualification reasons
✅ Verify the active segment list and every SID
✅ Confirm the BSID expands to the selected list
✅ Prove service routes consume the policy
✅ Fail the active candidate and verify backup convergence

The highest-preference valid candidate wins. A configured backup that is invalid provides no protection.

Continue the SR-MPLS Learning Journey:
https://www.youtube.com/playlist?list=PLYQHPRJxy5xk

Share your videos with friends, family, and the world

11/10/2026

New article: Understanding SRGB and SRLB 📘

A reliable SR-MPLS design starts with a clear label allocation plan. Two reserved blocks sit underneath every SR design:

SRGB: the global block. Prefix-SIDs and Node-SIDs are advertised as an index, and label = SRGB base + index. Index 100 in 16000-23999 is label 16100.

SRLB: the local block. Manual Adjacency-SIDs and local Binding SIDs. The same label number can mean something different on another router.

Inside the article (3 pages, attached as images):
- Why the outgoing label uses the next hop's SRGB, with a worked R1 to R4 example
- What IS-IS advertises (TLV 242, SR-Capabilities and SR Local Block sub-TLVs)
- How to plan the SID index space before choosing a range
- How to budget SRLB labels on your busiest router
- A worked allocation plan: 500 routers growing to 1000
- Pre-deployment checks: platform limits, collisions, persistence, migration

What SRGB base do you run, and have you ever changed one on a live network?

🔔 Subscribe on YouTube for a new SR-MPLS lesson every day:
https://www.youtube.com/?sub_confirmation=1

📺 Full 22-day playlist:
https://www.youtube.com/playlist?list=PLYQHPRJxy5xk

09/10/2026

Why IS-IS Uses Wide Metrics for Segment Routing over MPLS

Understanding the protocol requirement behind metric-style wide

Segment Routing over MPLS (SR-MPLS) relies on IS-IS extensions to distribute Segment Identifiers (SIDs). Wide metric mode enables the extended reachability formats that carry these attributes. Understanding this relationship explains why metric-style wide appears in IS-IS SR-MPLS configurations, even when every link uses a small metric. [1][2]

What wide metric mode changes

IS-IS encodes routing information in Type-Length-Value (TLV) structures. Legacy narrow link metrics use 6 bits, with values from 0 to 63. Extended IS Reachability TLV 22 uses a 24-bit link metric, while Extended IPv4 Reachability TLV 135 uses a 32-bit prefix metric. These extended formats also support sub-TLVs that attach additional information to links and prefixes. [1]

How Segment Routing uses the extensions

For a typical single-topology IPv4 deployment, the relevant advertisements are:

• Prefix-SID: carried in a sub-TLV within TLV 135. It associates a segment with an IP prefix, commonly a router loopback used as a Node-SID.

• Adjacency-SID: carried in a sub-TLV within TLV 22. It identifies a segment associated with a particular adjacency.

• SR capabilities and SRGB: advertised separately through Router Capability TLV 242. The Segment Routing Global Block describes the label space available for global segments.

Legacy narrow reachability TLVs cannot carry these Prefix-SID and Adjacency-SID sub-TLVs. IPv6 and multi-topology deployments use additional reachability TLVs defined by the SR extensions. The essential requirement is support for the extended encodings. [2]

Configuration example

Cisco IOS XE includes wide metric mode in its IS-IS SR-MPLS configuration procedure: [3]

router isis CORE

metric-style wide

segment-routing mpls

This excerpt assumes the remaining IS-IS, interface, MPLS and Prefix-SID configuration is in place. Command placement and defaults vary by platform; an explicit wide-metric command may be unnecessary where the required formats are already enabled.

References

[1] RFC 5305 — IS-IS Extensions for Traffic Engineering

https://www.rfc-editor.org/rfc/rfc5305.html

[2] RFC 8667 — IS-IS Extensions for Segment Routing

https://www.rfc-editor.org/rfc/rfc8667.html

[3] Cisco — Segment Routing with IS-IS v4 Node SID

https://www.cisco.com/c/en/us/td/docs/routers/ios-xe/segment-routing/segment-routing-guide/m_seg-routing-isis.html

Segment Routing with IS-IS v4 Node SID

08/10/2026

Day 8 of the 22-day SR-MPLS Learning Journey is live 🎥

Yesterday we turned Segment Routing on under IS-IS. Today it's OSPF — and the interesting part isn't how different it is. It's how little changes.

This one is different from the last few days: everything in it comes from a real four-router IOS XR lab. The configuration, the commands, the traceroute output and the label values are what the routers actually produced.

R1 — R2 — R3 — R4 in a line, one OSPF area, the default SRGB everywhere. Index 4 means label 16004, on every router.

Three things OSPF does differently from IS-IS:

1. No metric-style wide. OSPF carries SR in Opaque LSAs that were already part of the protocol. Nothing to enable.
2. Forwarding is a separate command. "segment-routing mpls" advertises. "segment-routing forwarding mpls" installs what arrives. Miss the second one and your database looks perfect while nothing forwards.
3. The Prefix-SID lives under the interface, inside the area block.

And the proof, straight from the lab — the traceroute from R1 to R4 shows the same label, 16004, at every hop. Not three different labels. One. That's Day 2's arithmetic, visible on the wire.

Everything SR needs rides in a single LSA — Type 10, the area-scoped Opaque. The index is advertised. The label is computed locally. That's the whole mechanism.

17 minutes, full lesson 👇
https://youtu.be/Ugb3zzkQiL8

🔔 Subscribe on YouTube for a new lesson every day:
https://www.youtube.com/?sub_confirmation=1

👍 Like this page for daily lesson summaries:
https://www.facebook.com/ThePacketPathWithPartha

📺 Full 22-day playlist:
https://www.youtube.com/playlist?list=PLYQHPRJxy5xk

Share your videos with friends, family, and the world

08/10/2026

An SR policy can be up and healthy while service traffic completely bypasses it. The usual cause is a mismatch between the BGP Color Extended Community, route next hop and SR policy endpoint.

Quick color-based steering checklist:
✅ Confirm the service route carries the expected color
✅ Verify the BGP next hop matches the policy endpoint
✅ Check that color + endpoint select the intended SR policy
✅ Validate the active candidate path and segment list
✅ Verify the BSID, recursive next hop, LFIB and hardware entry
✅ Confirm the service route resolves through the policy
✅ Test path failure and compare counters, loss and reoptimization

Do not stop after proving the policy is up—prove that the intended routes are consuming it.

Continue the SR-MPLS Learning Journey:
https://www.youtube.com/playlist?list=PLYQHPRJxy5xk

Share your videos with friends, family, and the world

04/10/2026

A Binding SID (BSID) lets one label represent an entire Segment Routing policy. It reduces the label stack exposed to upstream routers, but the policy still has to resolve correctly on the node that owns the BSID.

Quick BSID troubleshooting checklist:
✅ Confirm the BSID is allocated on the intended head-end
✅ Verify it maps to the correct active SR policy
✅ Decode the expanded segment list and next-hop recursion
✅ Confirm upstream reachability to the BSID owner
✅ Count repair, entropy, transport and service labels against the MSD
✅ Compare the policy with the LFIB and hardware table
✅ Fail the preferred path and verify reoptimization

A programmed BSID does not prove that its expanded segment list can be forwarded.

Continue the SR-MPLS Learning Journey:
https://www.youtube.com/playlist?list=PLYQHPRJxy5xk

Share your videos with friends, family, and the world

02/10/2026

TI-LFA protects SR-MPLS traffic only when a valid repair path is calculated and installed before the failure.

Quick verification checklist:
✅ Confirm the required link or node protection
✅ Verify TI-LFA is enabled in the correct IGP scope
✅ Check that a backup next hop is installed
✅ Inspect the repair segment list and imposed labels
✅ Validate SRGB/SRLB and Adj-SID reachability
✅ Test a controlled failure and measure packet loss
✅ Check label-depth and hardware limits

Fast timers help convergence, but they cannot replace a missing repair path.

Continue the SR-MPLS Learning Journey:
https://www.youtube.com/playlist?list=PLYQHPRJxy5xk

Share your videos with friends, family, and the world

01/10/2026

Uneven traffic in an SR-MPLS network does not always mean the route or label is wrong. If one ECMP path or LAG member is hot while others are idle, inspect how the flow is being hashed.

Quick Entropy Label / ECMP checklist:
✅ Verify every ECMP and LAG member is up
✅ Compare per-member packets, bytes and drops
✅ Confirm Entropy Label Indicator (ELI) and Entropy Label support
✅ Inspect the imposed label stack at ingress
✅ Check the device's MPLS hash inputs
✅ Test forward and reverse traffic separately
✅ Validate label-depth and hardware limits

An Entropy Label gives transit routers a stable value for distributing MPLS flows without deep payload inspection.

Learn more from the SR-MPLS Learning Journey:
https://www.youtube.com/playlist?list=PLYQHPRJxy5xk

Share your videos with friends, family, and the world