<?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: Fast ext4 fsck times, revisited</title>
	<atom:link href="http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/feed/" rel="self" type="application/rss+xml" />
	<link>http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/</link>
	<description>Musings about Open Source, Linux, and Life by Theodore Tso</description>
	<lastBuildDate>Mon, 22 Feb 2010 22:39:59 -0500</lastBuildDate>
	<generator>http://wordpress.org/?v=2.8.6</generator>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
		<item>
		<title>By: tytso</title>
		<link>http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/comment-page-1/#comment-2624</link>
		<dc:creator>tytso</dc:creator>
		<pubDate>Thu, 02 Jul 2009 15:57:12 +0000</pubDate>
		<guid isPermaLink="false">http://thunk.org/tytso/blog/?p=311#comment-2624</guid>
		<description>mat,

I&#039;ve added e2croncheck script into the contrib directory so it&#039;ll be in the e2fsprogs sources in the next release.  You can grab a copy of it here:

http://e2fsprogs.git.sourceforge.net/git/gitweb.cgi?p=e2fsprogs;a=blob_plain;f=contrib/e2croncheck;hb=7ad8da364950fd21aceb3c0a5785fda245036650</description>
		<content:encoded><![CDATA[<p>mat,</p>
<p>I&#8217;ve added e2croncheck script into the contrib directory so it&#8217;ll be in the e2fsprogs sources in the next release.  You can grab a copy of it here:</p>
<p><a href="http://e2fsprogs.git.sourceforge.net/git/gitweb.cgi?p=e2fsprogs;a=blob_plain;f=contrib/e2croncheck;hb=7ad8da364950fd21aceb3c0a5785fda245036650" rel="nofollow">http://e2fsprogs.git.sourceforge.net/git/gitweb.cgi?p=e2fsprogs;a=blob_plain;f=contrib/e2croncheck;hb=7ad8da364950fd21aceb3c0a5785fda245036650</a></p>
]]></content:encoded>
	</item>
	<item>
		<title>By: mat</title>
		<link>http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/comment-page-1/#comment-2622</link>
		<dc:creator>mat</dc:creator>
		<pubDate>Thu, 02 Jul 2009 11:52:31 +0000</pubDate>
		<guid isPermaLink="false">http://thunk.org/tytso/blog/?p=311#comment-2622</guid>
		<description>I like your &quot;finally&quot; answer, just I have no Idea on how do that.... is there some example on how:

1) create a snapshot
2) check the snapshot
3)if anything went fine, update the last checked time on the base filesystem

I never found any script such this on web..... You have a laptop with ext4 so I 
swear if you can publish your, please......</description>
		<content:encoded><![CDATA[<p>I like your &#8220;finally&#8221; answer, just I have no Idea on how do that&#8230;. is there some example on how:</p>
<p>1) create a snapshot<br />
2) check the snapshot<br />
3)if anything went fine, update the last checked time on the base filesystem</p>
<p>I never found any script such this on web&#8230;.. You have a laptop with ext4 so I<br />
swear if you can publish your, please&#8230;&#8230;</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: tytso</title>
		<link>http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/comment-page-1/#comment-2619</link>
		<dc:creator>tytso</dc:creator>
		<pubDate>Tue, 30 Jun 2009 14:41:41 +0000</pubDate>
		<guid isPermaLink="false">http://thunk.org/tytso/blog/?p=311#comment-2619</guid>
		<description>mat,

First of all, with the new allocator, it was taking about 14 seconds to fsck a 32GB partition in question, not 30 seconds.   Secondly, how the fsck time scales to a 250GB partition very much depends on the file system; it&#039;s a function of the number of inodes, average size of the files, average size of the directories, etc.   So it won&#039;t necessarily be a linear scale; if you use the 250GB file system to store a large number of large video files, for example, the fsck time will take much less than if you use that 250GB file system to store a gargantuan number of very small files in a very large number of directories.

Third of all, with a laptop, I recommend using suspend-to-ram as much as possible; if you&#039;re going to be using your laptop within a few hours, or if you can keep it plugged in, why shut it down?   Just use suspend-to-ram, and that will tend to reduce the number of mounts, which in turn will reduce the number of times fsck might kick in.

Finally, I strongly recommend that you use LVM, and then periodically run a script (perhaps out of cron) which creates a snapshot, checks the snapshot, and if the file system is consistent, updates the last checked time on the base file system.   This will eliminate the need to run fsck at boot time, and it will in practice mean that your file system can be checked for any corruption induced by hardware errors, et. al, much more regularly.</description>
		<content:encoded><![CDATA[<p>mat,</p>
<p>First of all, with the new allocator, it was taking about 14 seconds to fsck a 32GB partition in question, not 30 seconds.   Secondly, how the fsck time scales to a 250GB partition very much depends on the file system; it&#8217;s a function of the number of inodes, average size of the files, average size of the directories, etc.   So it won&#8217;t necessarily be a linear scale; if you use the 250GB file system to store a large number of large video files, for example, the fsck time will take much less than if you use that 250GB file system to store a gargantuan number of very small files in a very large number of directories.</p>
<p>Third of all, with a laptop, I recommend using suspend-to-ram as much as possible; if you&#8217;re going to be using your laptop within a few hours, or if you can keep it plugged in, why shut it down?   Just use suspend-to-ram, and that will tend to reduce the number of mounts, which in turn will reduce the number of times fsck might kick in.</p>
<p>Finally, I strongly recommend that you use LVM, and then periodically run a script (perhaps out of cron) which creates a snapshot, checks the snapshot, and if the file system is consistent, updates the last checked time on the base file system.   This will eliminate the need to run fsck at boot time, and it will in practice mean that your file system can be checked for any corruption induced by hardware errors, et. al, much more regularly.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: mat</title>
		<link>http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/comment-page-1/#comment-2617</link>
		<dc:creator>mat</dc:creator>
		<pubDate>Tue, 30 Jun 2009 11:39:49 +0000</pubDate>
		<guid isPermaLink="false">http://thunk.org/tytso/blog/?p=311#comment-2617</guid>
		<description>I still have some doubts.... this means a 32 Gb partition needs  more than 30 
secs fsck at boot. I have 250 Gb space in my laptop thus about 8 times fsck = 4 minutes. 
Now just imagine what happens when I need to turn on my PC for a conference or some other critical and urgent task and I need to wait 4 minutes for fsck + about 30-40 secs of boot....</description>
		<content:encoded><![CDATA[<p>I still have some doubts&#8230;. this means a 32 Gb partition needs  more than 30<br />
secs fsck at boot. I have 250 Gb space in my laptop thus about 8 times fsck = 4 minutes.<br />
Now just imagine what happens when I need to turn on my PC for a conference or some other critical and urgent task and I need to wait 4 minutes for fsck + about 30-40 secs of boot&#8230;.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: VP</title>
		<link>http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/comment-page-1/#comment-2042</link>
		<dc:creator>VP</dc:creator>
		<pubDate>Fri, 13 Mar 2009 15:51:12 +0000</pubDate>
		<guid isPermaLink="false">http://thunk.org/tytso/blog/?p=311#comment-2042</guid>
		<description>For what it&#039;s worth, JkDefrag[1], a defragmenter for Windows, groups all directories at the start of the disk. Apparently, the start of the disk is (a bit to significantly) faster than the end, and since directories are by far the most accessed files, it makes sense to put &#039;em there. I suppose a similar approach might be a good idea for ext4, if you&#039;re going to group &#039;em all anyway.

[1]http://www.kessels.com/Jkdefrag/</description>
		<content:encoded><![CDATA[<p>For what it&#8217;s worth, JkDefrag[1], a defragmenter for Windows, groups all directories at the start of the disk. Apparently, the start of the disk is (a bit to significantly) faster than the end, and since directories are by far the most accessed files, it makes sense to put &#8216;em there. I suppose a similar approach might be a good idea for ext4, if you&#8217;re going to group &#8216;em all anyway.</p>
<p>[1]http://www.kessels.com/Jkdefrag/</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Anon</title>
		<link>http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/comment-page-1/#comment-1974</link>
		<dc:creator>Anon</dc:creator>
		<pubDate>Thu, 12 Mar 2009 10:25:28 +0000</pubDate>
		<guid isPermaLink="false">http://thunk.org/tytso/blog/?p=311#comment-1974</guid>
		<description>(Darn it I was too slow in my request and the cloud of uncertainty has been spread far and wide. Sorry Ted. Feel free to delete this and the previous comment)</description>
		<content:encoded><![CDATA[<p>(Darn it I was too slow in my request and the cloud of uncertainty has been spread far and wide. Sorry Ted. Feel free to delete this and the previous comment)</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Anon</title>
		<link>http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/comment-page-1/#comment-1968</link>
		<dc:creator>Anon</dc:creator>
		<pubDate>Wed, 11 Mar 2009 17:56:09 +0000</pubDate>
		<guid isPermaLink="false">http://thunk.org/tytso/blog/?p=311#comment-1968</guid>
		<description>(OFFTOPIC)

Do you reckon you have time to bash out a &quot;safe file writing article&quot;? It looks like https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/317781 is getting a lot of traffic...</description>
		<content:encoded><![CDATA[<p>(OFFTOPIC)</p>
<p>Do you reckon you have time to bash out a &#8220;safe file writing article&#8221;? It looks like <a href="https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/317781" rel="nofollow">https://bugs.edge.launchpad.net/ubuntu/+source/linux/+bug/317781</a> is getting a lot of traffic&#8230;</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: tytso</title>
		<link>http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/comment-page-1/#comment-1947</link>
		<dc:creator>tytso</dc:creator>
		<pubDate>Wed, 04 Mar 2009 22:41:10 +0000</pubDate>
		<guid isPermaLink="false">http://thunk.org/tytso/blog/?p=311#comment-1947</guid>
		<description>@2: Ionut,

Unless some major problems are found with it, it will probably be merged at the next merge window (i.e., after 2.6.29 is released).    At this point my plans are to make it the default allocator, so no, you won&#039;t have to do anything special once you are booting a kernel that has the new allocator merged.

Of course, to get the most value out of the allocator, you&#039;ll need to do a backup/reformat/restore pass, so that the directory blocks are concentrated together, etc.   But it shouldn&#039;t do any harm to use the new allocator on an existing ext4 or ext3 filesystem; you just won&#039;t see all of the benefits of the new allocator.</description>
		<content:encoded><![CDATA[<p>@2: Ionut,</p>
<p>Unless some major problems are found with it, it will probably be merged at the next merge window (i.e., after 2.6.29 is released).    At this point my plans are to make it the default allocator, so no, you won&#8217;t have to do anything special once you are booting a kernel that has the new allocator merged.</p>
<p>Of course, to get the most value out of the allocator, you&#8217;ll need to do a backup/reformat/restore pass, so that the directory blocks are concentrated together, etc.   But it shouldn&#8217;t do any harm to use the new allocator on an existing ext4 or ext3 filesystem; you just won&#8217;t see all of the benefits of the new allocator.</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: Ionut</title>
		<link>http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/comment-page-1/#comment-1939</link>
		<dc:creator>Ionut</dc:creator>
		<pubDate>Wed, 04 Mar 2009 17:34:23 +0000</pubDate>
		<guid isPermaLink="false">http://thunk.org/tytso/blog/?p=311#comment-1939</guid>
		<description>i just read your post and i want to congrats you for your great job.
i do have a little question and i want to inform you that i&#039;m pretty new at this things.
if these new allocator will be merged soon and i use ext4 on my partitions, after do i change my kernel version do i have do to something to use the new allocator instead of the old one?</description>
		<content:encoded><![CDATA[<p>i just read your post and i want to congrats you for your great job.<br />
i do have a little question and i want to inform you that i&#8217;m pretty new at this things.<br />
if these new allocator will be merged soon and i use ext4 on my partitions, after do i change my kernel version do i have do to something to use the new allocator instead of the old one?</p>
]]></content:encoded>
	</item>
	<item>
		<title>By: GoblinX Project &#187; GoblinX Newsletter, Issue 189 (03/01/2009)</title>
		<link>http://thunk.org/tytso/blog/2009/02/26/fast-ext4-fsck-times-revisited/comment-page-1/#comment-1909</link>
		<dc:creator>GoblinX Project &#187; GoblinX Newsletter, Issue 189 (03/01/2009)</dc:creator>
		<pubDate>Sun, 01 Mar 2009 12:04:00 +0000</pubDate>
		<guid isPermaLink="false">http://thunk.org/tytso/blog/?p=311#comment-1909</guid>
		<description>[...] Fast ext4 fsck times, revisited [...]</description>
		<content:encoded><![CDATA[<p>[...] Fast ext4 fsck times, revisited [...]</p>
]]></content:encoded>
	</item>
</channel>
</rss>
