Chris's Wiki :: blog/solaris/ZFSRecordsizeGrowth Commentshttps://utcc.utoronto.ca/~cks/space/blog/solaris/ZFSRecordsizeGrowth?atomcommentsDWiki2018-02-14T04:32:23ZRecent comments in Chris's Wiki :: blog/solaris/ZFSRecordsizeGrowth.By Chris Siebenmann on /blog/solaris/ZFSRecordsizeGrowthtag:CSpace:blog/solaris/ZFSRecordsizeGrowth:781b19ee6d4cd7d16fb1696c05814211796f567dChris Siebenmann<div class="wikitext"><p>Belatedly: I've now done this test with my ashift=12 pool on 4k advance
sector disks (as opposed to the ashift=12 pool on 512n drives). It too
reports that a 16896 byte file has a <code>dblk</code> record size of 16.5K and
so on, the same as the example in the article. So I think this section
of ZFS is fully disconnected from the underlying reported disk sector
size.</p>
</div>2018-02-14T04:32:23ZBy D. Ebdrup on /blog/solaris/ZFSRecordsizeGrowthtag:CSpace:blog/solaris/ZFSRecordsizeGrowth:3170f56e4c585321e59e0d30bc29f1b38768bf7bD. Ebdruphttp://blog.nullrouted.org<div class="wikitext"><p>I really enjoy these posts you're doing on the internals of ZFS, keep 'em up.</p>
<p>I do wonder if this behaviour that you've investigated changes with ashift=12 when you've got disks with 4k physical block size.
Is that something you'd consider checking up on?</p>
</div>2018-02-05T17:46:19Z