<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Suraj Naikawadi on Medium]]></title>
        <description><![CDATA[Stories by Suraj Naikawadi on Medium]]></description>
        <link>https://medium.com/@suraj.naikawadi93?source=rss-9578fca84813------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*EotkEwhXVI6DqFRf</url>
            <title>Stories by Suraj Naikawadi on Medium</title>
            <link>https://medium.com/@suraj.naikawadi93?source=rss-9578fca84813------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 04:42:54 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@suraj.naikawadi93/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Enterprise WAN Resilience]]></title>
            <link>https://medium.com/@suraj.naikawadi93/enterprise-wan-resilience-60a75d6a459b?source=rss-9578fca84813------2</link>
            <guid isPermaLink="false">https://medium.com/p/60a75d6a459b</guid>
            <category><![CDATA[bgp]]></category>
            <category><![CDATA[gre]]></category>
            <category><![CDATA[ospf]]></category>
            <category><![CDATA[network-architecture]]></category>
            <category><![CDATA[networking]]></category>
            <dc:creator><![CDATA[Suraj Naikawadi]]></dc:creator>
            <pubDate>Sun, 18 Oct 2026 09:41:53 GMT</pubDate>
            <atom:updated>2026-10-18T09:41:53.988Z</atom:updated>
            <content:encoded><![CDATA[<p><em>When BGP stays Established, does that mean your application is still reachable?</em></p><p><strong>A dual-ISP enterprise lab using eBGP, GRE, OSPF, and automatic WAN failover</strong></p><h3>The challenge</h3><p>Two ISP connections. Two GRE tunnels. Two routing protocols.</p><p>But what happens when the primary ISP experiences a backbone failure while its customer-facing BGP session remains Established?</p><p>Will the network detect the outage? Will application traffic automatically move to the secondary ISP? And how long will recovery take?</p><p>These are the questions I wanted to explore with <strong>Enterprise WAN Resilience</strong>, a lab built using the PingMyNetwork Sandbox. <a href="https://pingmynetwork.com/sandbox">https://pingmynetwork.com/sandbox</a></p><p>In enterprise network design, redundancy is more than adding another WAN connection. A resilient network needs to detect failures, select an available path, and restore application connectivity without manual intervention.</p><p>In this lab, we separate the network into two routing layers:</p><ul><li><strong>Underlay:</strong> eBGP and ISP-internal OSPF provide WAN connectivity.</li><li><strong>Overlay:</strong> GRE tunnels carry enterprise OSPF routes between Headquarters and the Disaster Recovery site.</li></ul><p>Let’s build the network and see how these layers work together.</p><h3>1. Network topology and scenario</h3><p>Our fictional organization operates two locations:</p><p><strong>Headquarters (HQ)</strong> — Primary enterprise location.</p><p><strong>Disaster Recovery (DR)</strong> — Secondary site hosting application services.</p><p>Both sites connect to two independent ISPs.</p><ul><li><strong>ISP-A (AS 65100)</strong> is the preferred transport path.</li><li><strong>ISP-B (AS 65200)</strong> provides an independent backup path.</li><li><strong>HQ-CE (AS 65010)</strong> and <strong>DR-CE (AS 65020)</strong> establish eBGP sessions to their respective providers.</li><li><strong>GRE Tunnel100</strong> traverses ISP-A; <strong>GRE Tunnel200</strong> traverses ISP-B.</li><li><strong>Enterprise OSPF process 10, area 0</strong> runs across both GRE tunnels and advertises the LANs.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eHU3lJg-ZaaXnw8qX6J0qg.png" /><figcaption>Figure 1 — Physical ISP networks and logical GRE overlays</figcaption></figure><p>The lab has <strong>six routers</strong>, two site switches, and client &amp; server test nodes. The ISP routers act as a simulated provider network; the enterprise routers own the GRE tunnels and internal OSPF process.</p><p><strong>Design principle:</strong> We don’t redistribute enterprise OSPF routes into Internet-facing eBGP. The providers carry only the routes needed to reach the remote GRE endpoints.</p><h3>Addressing plan</h3><p>WAN access and provider backbones</p><p><strong>ISP-A — AS 65100 (primary provider)</strong></p><ul><li><strong>HQ-CE ↔ ISP-A-HQ:</strong> 192.0.2.1/30 ↔ 192.0.2.2/30 (eBGP AS 65010 ↔ 65100)</li><li><strong>ISP-A backbone:</strong> 10.255.0.1/30 ↔ 10.255.0.2/30 (provider OSPF process 100)</li><li><strong>ISP-A-DR ↔ DR-CE:</strong> 192.0.2.6/30 ↔ 192.0.2.5/30 (eBGP AS 65100 ↔ 65020)</li></ul><p><strong>ISP-B — AS 65200 (secondary provider)</strong></p><ul><li><strong>HQ-CE ↔ ISP-B-HQ:</strong> 192.0.2.9/30 ↔ 192.0.2.10/30 (eBGP AS 65010 ↔ 65200)</li><li><strong>ISP-B backbone:</strong> 10.255.0.5/30 ↔ 10.255.0.6/30 (provider OSPF process 200)</li><li><strong>ISP-B-DR ↔ DR-CE:</strong> 192.0.2.14/30 ↔ 192.0.2.13/30 (eBGP AS 65200 ↔ 65020)</li></ul><h3>GRE overlay</h3><p><strong>Tunnel100 — primary path across ISP-A</strong></p><ul><li>HQ GRE address: 192.168.100.1/30; DR GRE address: 192.168.100.2/30</li><li>GRE outer endpoints: 192.0.2.1 (HQ) → 192.0.2.5 (DR)</li><li>OSPF: Area 0; configured tunnel bandwidth: <strong>100,000 Kbps</strong>; expected OSPF interface cost: <strong>10</strong></li></ul><p><strong>Tunnel200 — backup path across ISP-B</strong></p><ul><li>HQ GRE address: 192.168.200.1/30; DR GRE address: 192.168.200.2/30</li><li>GRE outer endpoints: 192.0.2.9 (HQ) → 192.0.2.13 (DR)</li><li>OSPF: Area 0; configured tunnel bandwidth: <strong>10,000 Kbps</strong>; expected OSPF interface cost: <strong>100</strong></li></ul><h3>Site networks</h3><p><strong>Headquarters</strong></p><ul><li><strong>VLAN 10 — Users:</strong> 10.10.10.0/24, default gateway 10.10.10.1</li><li><strong>VLAN 20 — Servers:</strong> 10.10.20.0/24, default gateway 10.10.20.1</li></ul><p><strong>Disaster Recovery site</strong></p><ul><li><strong>VLAN 10 — Users:</strong> 10.20.10.0/24, default gateway 10.20.10.1</li><li><strong>VLAN 20 — Servers:</strong> 10.20.20.0/24, default gateway 10.20.20.1</li></ul><h3>Phase 1 — Make the ISP underlay work first</h3><p>The goal of phase 1 is straightforward: <strong>HQ must be able to reach the DR router’s WAN address through each provider, without involving GRE or enterprise OSPF.</strong></p><p>Each ISP has two edge routers. Its internal OSPF process advertises the locally connected CE-facing /30 across the backbone. The opposite provider edge router learns that remote /30, then advertises it to its customer through eBGP.</p><h3>Configure the ISP-A routers</h3><p>On <strong>ISP-A-HQ</strong>, the important provider-side routing statements are:</p><pre>router ospf 100<br> passive-interface default<br> no passive-interface GigabitEthernet0/1<br> network 10.255.0.0 0.0.0.3 area 0<br> network 192.0.2.0 0.0.0.3 area 0<br>!<br>ip prefix-list WAN-TO-CE-OUT seq 5 permit 192.0.2.4/30<br>router bgp 65100<br> neighbor 192.0.2.1 remote-as 65010<br> neighbor 192.0.2.1 prefix-list WAN-TO-CE-OUT out<br> network 192.0.2.4 mask 255.255.255.252</pre><p>On <strong>ISP-A-DR</strong>, mirror the logic for the other access subnet:</p><pre>router ospf 100<br> passive-interface default<br> no passive-interface GigabitEthernet0/1<br> network 10.255.0.0 0.0.0.3 area 0<br> network 192.0.2.4 0.0.0.3 area 0<br>!<br>ip prefix-list WAN-TO-CE-OUT seq 5 permit 192.0.2.0/30<br>router bgp 65100<br> neighbor 192.0.2.5 remote-as 65020<br> neighbor 192.0.2.5 prefix-list WAN-TO-CE-OUT out<br> network 192.0.2.0 mask 255.255.255.252</pre><p>We apply the same pattern to ISP-B with <strong>OSPF process 200</strong> and <strong>AS 65200</strong>. ISP-B-HQ advertises the remote 192.0.2.12/30 to HQ; ISP-B-DR advertises 192.0.2.8/30 to DR. The full six-router configurations are available in the accompanying lab configuration appendix.</p><p><strong>A useful detail:</strong> The customer-facing interface is <em>passive</em> in the provider’s OSPF process. This allows the ISP to advertise the connected /30 internally without attempting an OSPF adjacency with the customer CE. Only the two routers inside a provider become OSPF neighbors.</p><p>The BGP network command originates the remote /30 only when an exact matching route exists in the local routing table. That dependency becomes valuable during the failure test.</p><h3>Establish eBGP at the enterprise edges</h3><p>On <strong>HQ-CE</strong>:</p><pre>ip prefix-list ISP-A-IN seq 10 permit 192.0.2.4/30<br>ip prefix-list ISP-B-IN seq 10 permit 192.0.2.12/30<br>!<br>router bgp 65010<br> bgp router-id 10.255.255.1<br> neighbor 192.0.2.2 remote-as 65100<br> neighbor 192.0.2.2 prefix-list ISP-A-IN in<br> neighbor 192.0.2.10 remote-as 65200<br> neighbor 192.0.2.10 prefix-list ISP-B-IN in</pre><p>On <strong>DR-CE</strong>:</p><pre>ip prefix-list ISP-A-IN seq 10 permit 192.0.2.0/30<br>ip prefix-list ISP-B-IN seq 10 permit 192.0.2.8/30<br>!<br>router bgp 65020<br> bgp router-id 10.255.255.2<br> neighbor 192.0.2.6 remote-as 65100<br> neighbor 192.0.2.6 prefix-list ISP-A-IN in<br> neighbor 192.0.2.14 remote-as 65200<br> neighbor 192.0.2.14 prefix-list ISP-B-IN in</pre><h3>Verify the underlay</h3><p>On HQ-CE:</p><pre>show ip bgp summary<br>show ip route 192.0.2.5<br>show ip route 192.0.2.13<br>ping 192.0.2.5 source 192.0.2.1<br>ping 192.0.2.13 source 192.0.2.9</pre><p>Expected route selection on HQ:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/997/1*68jmL7KlkBD3HJqCI2w9BQ.png" /><figcaption><strong>Screenshot 1 — </strong>show ip bgp summary on HQ-CE</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/939/1*nKXea_r_btWe-5Lyd7z-Wg.png" /><figcaption><strong>Screenshot 2 — </strong>WAN routes on HQ-CE</figcaption></figure><h3>Phase 2 — Build two GRE tunnels</h3><p>With the WAN underlay reachable, GRE gives us stable point-to-point interfaces between HQ and DR. Unlike a basic routed WAN connection, GRE can carry the OSPF multicast packets required to discover neighbors over the overlay.</p><h3>HQ-CE — GRE interfaces</h3><pre>interface Tunnel100<br> description PRIMARY-GRE-VIA-ISP-A<br> ip address 192.168.100.1 255.255.255.252<br> bandwidth 100000<br> tunnel source 192.0.2.1<br> tunnel destination 192.0.2.5<br>!<br>interface Tunnel200<br> description BACKUP-GRE-VIA-ISP-B<br> ip address 192.168.200.1 255.255.255.252<br> bandwidth 10000<br> tunnel source 192.0.2.9<br> tunnel destination 192.0.2.13</pre><h3>DR-CE — GRE interfaces</h3><pre>interface Tunnel100<br> description PRIMARY-GRE-VIA-ISP-A<br> ip address 192.168.100.2 255.255.255.252<br> bandwidth 100000<br> tunnel source 192.0.2.5<br> tunnel destination 192.0.2.1<br>!<br>interface Tunnel200<br> description BACKUP-GRE-VIA-ISP-B<br> ip address 192.168.200.2 255.255.255.252<br> bandwidth 10000<br> tunnel source 192.0.2.13<br> tunnel destination 192.0.2.9</pre><p>Notice that <strong>the tunnels have different transport source/destination IPs</strong>. This is intentional: Tunnel100 follows ISP-A, and Tunnel200 follows ISP-B. A tunnel is only useful as an independent backup if its destination route stays tied to its intended provider.</p><p>Verify from HQ:</p><pre>show ip interface brief<br>ping 192.168.100.2 source 192.168.100.1<br>ping 192.168.200.2 source 192.168.200.1</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/963/1*-5-vSJV9lzvrl0wGhiTmtQ.png" /><figcaption><strong>Screenshot 3— </strong>GRE verification</figcaption></figure><h3>Phase 3 — Use OSPF for the enterprise overlay</h3><p>Now we enable enterprise OSPF. This OSPF process runs <strong>between the sites</strong>, not with the ISPs. Internal LAN prefixes are exchanged across both GRE tunnels.</p><h3>HQ-CE</h3><pre>router ospf 10<br> router-id 10.255.255.1<br> auto-cost reference-bandwidth 1000<br> passive-interface default<br> no passive-interface Tunnel100<br> no passive-interface Tunnel200<br> network 10.10.10.0 0.0.0.255 area 0<br> network 10.10.20.0 0.0.0.255 area 0<br> network 192.168.100.0 0.0.0.3 area 0<br> network 192.168.200.0 0.0.0.3 area 0</pre><h3>DR-CE</h3><pre>router ospf 10<br> router-id 10.255.255.2<br> auto-cost reference-bandwidth 1000<br> passive-interface default<br> no passive-interface Tunnel100<br> no passive-interface Tunnel200<br> network 10.20.10.0 0.0.0.255 area 0<br> network 10.20.20.0 0.0.0.255 area 0<br> network 192.168.100.0 0.0.0.3 area 0<br> network 192.168.200.0 0.0.0.3 area 0</pre><p>passive-interface default prevents OSPF Hellos from being sent toward users and servers. The two no passive-interface lines allow OSPF to form a neighbor relationship across <strong>both</strong> GRE tunnels.</p><h3>Why Tunnel100 is preferred</h3><p>Both tunnel interfaces should become OSPF neighbors when healthy. Path preference is a separate decision, based on cost.</p><p>We set the OSPF reference bandwidth to <strong>1,000 Mbps</strong> and the logical tunnel bandwidths to <strong>100 Mbps</strong> and <strong>10 Mbps</strong>:</p><pre>Tunnel100 cost = 1000 / 100 = 10<br>Tunnel200 cost = 1000 / 10  = 100</pre><p>These bandwidth values influence the routing metric; they don’t enforce actual tunnel speed. Assuming otherwise equal destination-side costs, HQ installs the route to DR’s LAN via <strong>Tunnel100</strong>, while Tunnel200 remains available as an alternative.</p><p>On HQ-CE:</p><pre>show ip ospf neighbor<br>show ip ospf interface Tunnel100<br>show ip ospf interface Tunnel200<br>show ip route ospf<br>show ip route 10.20.20.0</pre><p>Expected healthy state:</p><ul><li>Neighbor 10.255.255.2 is <strong>FULL</strong> across Tunnel100.</li><li>The same neighbor is <strong>FULL</strong> across Tunnel200.</li><li>The selected DR LAN route uses 192.168.100.2 through Tunnel100.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/897/1*5tBz8C7_4ujNDHxD2sg1DQ.png" /><figcaption><strong>Screenshot 4 — </strong>OSPF adjacencies.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/919/1*ApSIDoftXfT1v6IWj-2Axg.png" /><figcaption><strong>Screenshot 5 — </strong>OSPF process on Tunnel interfaces.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/912/1*uriYdDGPdq4xsv76pej8yA.png" /><figcaption><strong>Screenshot 6— </strong>OSPF learned routes &amp; preferred OSPF route through Tunnel100.</figcaption></figure><h3>Phase 4 — The WAN resilience challenge</h3><p>Here’s the part I find most interesting: <strong>What happens when the primary ISP fails somewhere between its two edge routers, rather than at the customer’s local WAN handoff?</strong></p><p>Our ISP-A routers are connected across 10.255.0.0/30. We can simulate a provider backbone outage by shutting the ISP-A core-facing interface on <strong>ISP-A-HQ</strong>. This intentionally breaks the ISP-A internal OSPF adjacency but leaves the customer-facing BGP links physically connected.</p><h3>Baseline checks</h3><p>Before introducing a fault, capture these commands:</p><pre>! HQ-CE<br>show ip bgp summary<br>show ip ospf neighbor<br>show ip route 10.20.20.0<br>ping 10.20.20.1 repeat 100</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/997/1*gGAi3msUWR6mx8m6EHFfow.png" /><figcaption><strong>Screenshot 7</strong> — BGP and OSPF Neighbor status.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/850/1*Xb57siaYBhkf3MH1cVK8Qg.png" /><figcaption><strong>Screenshot 8 </strong>— SRV2 Webpage accessible from PC1.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/639/1*8O-GK6rJZ-Ng_ssgPco_IA.png" /><figcaption><strong>Screenshot 9 </strong>— Ping result from PC1 to PC2.</figcaption></figure><h3>Introduce the ISP-A core failure</h3><p>On <strong>ISP-A-HQ</strong>:</p><pre>configure terminal<br>interface GigabitEthernet0/1<br> shutdown<br>end</pre><p>Then check HQ-CE:</p><pre>show ip bgp summary<br>show ip route 192.0.2.5<br>show ip ospf neighbor<br>show ip route 10.20.20.0</pre><h3>What should happen, and why</h3><ol><li>The <strong>HQ-CE ↔ ISP-A-HQ eBGP session may remain Established</strong>, because that local CE-to-ISP access link is still up.</li><li>ISP-A’s internal OSPF route to the far-end WAN subnet 192.0.2.4/30 disappears when its core fails.</li><li>The ISP-A-HQ BGP network 192.0.2.4/30 advertisement is withdrawn once the exact route no longer exists in its routing table.</li><li>The primary GRE destination 192.0.2.5 becomes unreachable through ISP-A. The OSPF adjacency over Tunnel100 is then lost as the failure is detected.</li><li>Enterprise traffic can switch to the surviving route through <strong>Tunnel200 / ISP-B</strong>, provided the backup GRE tunnel and its OSPF adjacency remain healthy.</li></ol><p>This is the difference between a routing protocol being <strong>up locally</strong> and a service being <strong>reachable end to end</strong>. The precise outage duration depends on routing withdrawals and OSPF/GRE failure-detection behavior in the simulator; we should measure it rather than assume instant convergence.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wOqPzBprNTU21b81xRhVRA.png" /><figcaption><strong>Screenshot 10</strong> — the unchanged eBGP session state.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/876/1*PRhMBi9HinKQm1OLl70NKg.png" /><figcaption><strong>Screenshot 11</strong> — the lost Tunnel100 neighbor.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/758/1*ji2eB7dJcEbBRQC79n3CbQ.png" /><figcaption><strong>Screenshot 12</strong> — the route switching to 192.168.200.2</figcaption></figure><h3>Restore ISP-A</h3><p>On ISP-A-HQ:</p><pre>configure terminal<br>interface GigabitEthernet0/1<br> no shutdown<br>end</pre><p>Verify that provider OSPF and BGP route propagation recover. Once both enterprise OSPF adjacencies are FULL again, the lower-cost Tunnel100 path should become preferred.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pFkWuAYJh44dIo0XR4WLDg.png" /><figcaption><strong>Screenshot 13 </strong>— Restore ISP-A</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/742/1*akppGsl10SaTYVOsjBfXiA.png" /><figcaption><strong>Screenshot 14</strong> — Tunnel100 path restored.</figcaption></figure><h3>Phase 5 — Measure the application impact</h3><p>A topology that recovers is good. A topology whose recovery we can <strong>measure</strong> is more useful.</p><p>For a basic test, continuously ping the PC2 10.20.10.10 from PC1 while reproducing the provider outage.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/690/1*mILstoMc7aU6Fg52mnzbvQ.png" /><figcaption><strong>Screenshot 15</strong> — Ping result from PC1 to PC2. Ping loss during outage and traffic shift to backup path through Tunnel200.</figcaption></figure><h3>Lessons from the architecture</h3><p><strong>1. Two ISPs do not automatically mean resilience.</strong> Each ISP needs an independently reachable path, and the enterprise needs a method to detect and avoid failures.</p><p><strong>2. Underlay and overlay solve different problems.</strong> The providers use eBGP and internal OSPF to transport the GRE endpoints. The enterprise runs its own OSPF routing domain across GRE to reach private subnets.</p><p><strong>3. OSPF cost selects a path, not a neighbor.</strong> Both GRE neighbors can be FULL simultaneously while the lower-cost Tunnel100 handles normal traffic.</p><p><strong>4. An Established BGP session is not an application-health check.</strong> The local peering can survive while remote reachability fails due to a problem deeper inside the provider network.</p><p><strong>5. GRE provides encapsulation, not confidentiality.</strong> It makes the overlay and OSPF multicast routing possible, but GRE traffic is not encrypted. For real traffic over an untrusted WAN, <strong>GRE over IPsec</strong> or another secure, authenticated transport is a natural security enhancement, subject to platform support and design validation.</p><p><em>Built in the PingMyNetwork Sandbox as part of the Enterprise WAN Resilience lab challenge.</em></p><p><strong>Tags:</strong> #Networking #BGP #OSPF #GRE #NetworkArchitecture</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=60a75d6a459b" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>