<p>Live gcvis: <a href="http://catbag.frankerfacez.com:8081/">http://catbag.frankerfacez.com:8081/</a> </p>
<p>Screenshot: <a href="http://i.imgur.com/GuORPBw.png">http://i.imgur.com/GuORPBw.png</a></p>
<p>Allocations by size bucket: <a href="http://putco.de/MjM3Mg.smalltalk">http://putco.de/MjM3Mg.smalltalk</a></p>
<p>Client count: 3,530</p>
<p>Why is the RSS so much higher than the gc.heapinuse? It looks like a leak, but it doesn't look like the last few leaks I fixed (adding huge objects to an array and never clearing it, etc).</p>
<hr/>**评论：**<br/><br/>matttproud: <pre><p>Without knowing 1. whether you have fixed all the leaks and 2. the lifecycle of your allocations, it is worth mentioning that the Go runtime's garbage collector is is <a href="https://en.wikipedia.org/wiki/Tracing_garbage_collection">non-moving</a>, which implies there's no heap compaction. I am not saying that heap fragmentation is occurring in your case, but it could be a possibility.</p>
<p>I have an <a href="https://github.com/golang/go/issues/11890">open bug</a> for a feature request to acquire more telemetric data from the runtime that would be useful in situations like these.</p></pre>riking27: <pre><p>Actually, heap fragmentation is a really likely answer.</p></pre>djherbis: <pre><p>What version of Go are you running?</p></pre>
