<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Ssh on Matteo Croce personal homepage</title><link>https://matteocroce.it/tags/ssh/</link><description>Recent content in Ssh on Matteo Croce personal homepage</description><generator>Hugo</generator><language>en-US</language><lastBuildDate>Thu, 16 Apr 2026 01:21:14 +0200</lastBuildDate><atom:link href="https://matteocroce.it/tags/ssh/index.xml" rel="self" type="application/rss+xml"/><item><title>Low Quality of Service: How My Router Broke Git</title><link>https://matteocroce.it/blog/ssh-low-qos/</link><pubDate>Thu, 16 Apr 2026 01:21:14 +0200</pubDate><guid>https://matteocroce.it/blog/ssh-low-qos/</guid><description>&lt;p&gt;QoS stands for Quality of Service. Improvement is not implied.&lt;/p&gt;
&lt;h2 id="symptoms"&gt;Symptoms&lt;/h2&gt;
&lt;p&gt;I&amp;rsquo;ve been a happy OpenWrt user for 20 years,
but due to a hardware failure I had to temporarily switch back to my ISP router.&lt;br&gt;
After the swap, every &lt;code&gt;git pull&lt;/code&gt; started hanging.
No errors, just a timeout after a few minutes.&lt;/p&gt;
&lt;h2 id="the-investigation"&gt;The Investigation&lt;/h2&gt;
&lt;p&gt;The usual suspects were checked first:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;DNS&lt;/strong&gt; - always the first suspect, but manual resolution with &lt;code&gt;dig&lt;/code&gt; worked fine.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;MTU/MSS clamping&lt;/strong&gt; - second suspect, but usually also breaks HTTP downloads.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;IPv6&lt;/strong&gt; - I disabled IPv6 just in case, but the problem persisted.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Firewall rules&lt;/strong&gt; - I could connect with netcat to github.com:22 and see the SSH banner.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;I didn&amp;rsquo;t find any issues with general connectivity,
so I focused on Git and noticed that some repositories were working while others weren&amp;rsquo;t.
Initially I thought that it could be a server-side issue,
but the failing repositories were spread between GitHub and GitLab,
so it was unlikely that both services had simultaneous issues.&lt;/p&gt;</description></item></channel></rss>