<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Umesh Racha – Insights</title>
    <link>https://umeshracha.com/insights</link>
    <description>Practical notes on ISP and media platforms: TR-069/GenieACS, CAS and OSS/BSS integrations, and high-availability architecture.</description>
    <language>en</language>
    <managingEditor>iam@umeshracha.com (Umesh Racha)</managingEditor>
    <webMaster>iam@umeshracha.com (Umesh Racha)</webMaster>
    <lastBuildDate>Mon, 26 Oct 2026 00:00:00 GMT</lastBuildDate>
    <atom:link href="https://umeshracha.com/rss.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>HAProxy + keepalived for RADIUS and billing: the topology that stops 3 a.m. outages</title>
      <link>https://umeshracha.com/insights/haproxy-keepalived-for-radius-and-billing</link>
      <guid isPermaLink="true">https://umeshracha.com/insights/haproxy-keepalived-for-radius-and-billing</guid>
      <pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate>
      <description>RADIUS is UDP, billing is HTTP and the database is a cluster. One pair of front-end nodes running keepalived and HAProxy can keep all three online if you split the roles correctly. The topology, the checks and the failure drills.</description>
      <category>HAProxy</category>
      <category>keepalived</category>
      <category>RADIUS</category>
      <category>High availability</category>
    </item>
    <item>
      <title>Moving CPE fleets from a vendor ACS to GenieACS without a support storm</title>
      <link>https://umeshracha.com/insights/moving-cpe-fleets-from-vendor-acs-to-genieacs</link>
      <guid isPermaLink="true">https://umeshracha.com/insights/moving-cpe-fleets-from-vendor-acs-to-genieacs</guid>
      <pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate>
      <description>A staged, checklist-driven way to move routers and ONTs from a vendor ACS to GenieACS: inventory, presets per model, batch-wise ACS URL changes, verification gates and a rollback that is just a URL.</description>
      <category>TR-069</category>
      <category>GenieACS</category>
      <category>CPE</category>
      <category>Migration</category>
    </item>
    <item>
      <title>Why CAS entitlements drift — and the reconciliation loop that fixes it</title>
      <link>https://umeshracha.com/insights/why-cas-entitlements-drift-and-the-reconciliation-loop</link>
      <guid isPermaLink="true">https://umeshracha.com/insights/why-cas-entitlements-drift-and-the-reconciliation-loop</guid>
      <pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate>
      <description>Entitlement drift between your subscriber management system and the CAS is a retry and reconciliation problem, not a vendor problem. Here is where drift comes from and the loop that keeps Nagra, SafeView or ABV in step with billing.</description>
      <category>CAS</category>
      <category>IPTV</category>
      <category>OSS/BSS</category>
      <category>Reconciliation</category>
    </item>
  </channel>
</rss>
