<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
		>
<channel>
	<title>Comments on: Is ext4 unsafe?</title>
	<atom:link href="http://tribulaciones.org/2009/03/is-ext4-unsafe/feed/" rel="self" type="application/rss+xml" />
	<link>http://tribulaciones.org/2009/03/is-ext4-unsafe/</link>
	<description>A blog about me and whatever insterests me</description>
	<lastBuildDate>Mon, 06 Apr 2009 12:59:07 -0700</lastBuildDate>
	<generator>http://wordpress.org/?v=2.8.4</generator>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
		<item>
		<title>By: ssam</title>
		<link>http://tribulaciones.org/2009/03/is-ext4-unsafe/comment-page-1/#comment-45</link>
		<dc:creator>ssam</dc:creator>
		<pubDate>Fri, 20 Mar 2009 12:26:11 +0000</pubDate>
		<guid isPermaLink="false">http://tribulaciones.org/?p=115#comment-45</guid>
		<description>if you dont care about the state of the filesystem after a crash/power outage, why use a filesystem with journalling, and why bother having fsck installed?

but what if you want to update a file every few minutes/seconds, (history file in a browser, prefs file etc). you dont want to actually write to disk straight away (for performance, powersaving, and ssd lifetime). in the event of a system crash i dont mind if i loose the last half hour of updates to those files, but i will cry if the files are completely deleted.

open(foo.new)
write(foo.new)
close(foo.new)
rename(foo.new, foo)

seems a good way of expressing it. all it demands from the file system is that the rename happens after the actual write of foo.new.

ext4 is currently ruining peoples days because it delays the data write (a good thing), but very quick writes the rename.

the patches makes the data write very quick, so you loose all the advantages of delaying the write.

another solution would be to delay the rename until after the data hits the disk.

another solution, might be to remember the blocks on which the old data was. in the event of a crash the filename could be pointed back to the old data.</description>
		<content:encoded><![CDATA[<p>if you dont care about the state of the filesystem after a crash/power outage, why use a filesystem with journalling, and why bother having fsck installed?</p>
<p>but what if you want to update a file every few minutes/seconds, (history file in a browser, prefs file etc). you dont want to actually write to disk straight away (for performance, powersaving, and ssd lifetime). in the event of a system crash i dont mind if i loose the last half hour of updates to those files, but i will cry if the files are completely deleted.</p>
<p>open(foo.new)<br />
write(foo.new)<br />
close(foo.new)<br />
rename(foo.new, foo)</p>
<p>seems a good way of expressing it. all it demands from the file system is that the rename happens after the actual write of foo.new.</p>
<p>ext4 is currently ruining peoples days because it delays the data write (a good thing), but very quick writes the rename.</p>
<p>the patches makes the data write very quick, so you loose all the advantages of delaying the write.</p>
<p>another solution would be to delay the rename until after the data hits the disk.</p>
<p>another solution, might be to remember the blocks on which the old data was. in the event of a crash the filename could be pointed back to the old data.</p>
]]></content:encoded>
	</item>
</channel>
</rss>

<!-- www.000webhost.com Analytics Code -->
<script type="text/javascript" src="http://analytics.hosting24.com/count.php"></script>
<noscript><a href="http://www.hosting24.com/"><img src="http://analytics.hosting24.com/count.php" alt="web hosting" /></a></noscript>
<!-- End Of Analytics Code -->
