<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Nftables on Marcello Barnaba</title>
    <link>https://sindro.me/tags/nftables/</link>
    <description>Recent content in Nftables on Marcello Barnaba</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Fri, 01 May 2026 00:00:00 +0000</lastBuildDate>
    <atom:link href="https://sindro.me/tags/nftables/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>mwan3 Failover Without the Hung Connections</title>
      <link>https://sindro.me/posts/2026-05-01-mwan3-failover-conntrack/</link>
      <pubDate>Fri, 01 May 2026 00:00:00 +0000</pubDate>
      <guid>https://sindro.me/posts/2026-05-01-mwan3-failover-conntrack/</guid>
      <description>&lt;p&gt;&lt;strong&gt;TL;DR:&lt;/strong&gt; &lt;a href=&#34;https://openwrt.org/docs/guide-user/network/wan/multiwan/mwan3&#34; target=&#34;_blank&#34;&gt;mwan3&lt;/a&gt;&#xA;reroutes &lt;em&gt;new&lt;/em&gt; flows when an uplink dies. Existing flows stay pinned to&#xA;the dead path — conntrack remembers, the firewall flow offload keeps&#xA;shovelling packets along it, and long-lived TCP sockets linger until&#xA;their application notices and reconnects. The native &lt;code&gt;flush_conntrack&lt;/code&gt;&#xA;option is a global nuke. The fix is a fifteen-line &lt;code&gt;/etc/mwan3.user&lt;/code&gt;&#xA;that does a &lt;em&gt;selective&lt;/em&gt; conntrack flush by mwan3 mark on &lt;code&gt;disconnected&lt;/code&gt;&#xA;events only.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
