linux-btrfs.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: George Mitchell <george@chinilu.com>
To: "Szőts Ákos" <szotsaki@gmail.com>
Cc: linux-btrfs@vger.kernel.org
Subject: Re: systemd "journalctl" is 1.89sec on ext4, 1.49 min on btrfs
Date: Mon, 27 May 2013 09:42:13 -0700	[thread overview]
Message-ID: <51A38CE5.1010003@chinilu.com> (raw)
In-Reply-To: <16735484.PAleCPYpY8@noname>

I have gotten what appear to be large increases in speed out of btrfs by 
defragmentation of meta data. The manual defragmentation process takes 
forever as you have to defragment incrementally directory by directory.  
I was at the point where KDE startup times were getting abysmal (along 
with journalctl, etc) and the multiple drives would churn incessantly on 
startup.  In the case of KDE, I found almost magical improvement with 
one operation:  `btrfs filesystem defrag /usr/share`.  I am currently 
going through the whole system deframenting directory by directory.  Its 
amazing, it proceeds quite quickly and then hits a directory at random 
where it sits and plods away seemingly forever before moving on.  I am 
convinced that there is something going on here with meta data 
fragmentation that, at times, is seriously affecting performance.  I 
*think* that autodefrag, once its out the door will hopefully solve 
this, in the mean time I am trying to come up with some sort of way to 
schedule an anacron job to deal with this issue.  But my suggestion 
would be that you try defragging your /var filesystem as thoroughly as 
possible on the meta data side.

On 05/27/2013 09:21 AM, Szőts Ákos wrote:
> Dear list,
>
> I have two openSUSE 12.3 systems with kernel 3.9. On one of them there's an
> ext4 partition, while on the other there's a btrfs.
>
> I issued a "time journalctl -b --no-pager" command on both systems. This shows
> the logs from the current boot without passing them to "less".
>
> On ext4 (3.9.3):
> real    0m1.898s
> user    0m0.291s
> sys     0m0.105s
>
> On btrfs (3.9.2):
> real    1m49.698s
> user    0m0.102s
> sys     0m0.470s
>
> Journalctl on btrfs was always this slow, some btrfsck were made on the file
> system too, but I don't think it was corrupted. On just the first run it's
> sluggish, after it's fast as the ext4 one.
>
> Is it a known issue or can I help somehow debugging this further?
>
> Best regards,
>
> Ákos Szőts
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>


  parent reply	other threads:[~2013-05-27 16:42 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-05-27 16:21 systemd "journalctl" is 1.89sec on ext4, 1.49 min on btrfs Szőts Ákos
     [not found] ` <CAFvQSYTtuH99nqMb6589xJgwsrSG6fD02XzU3ji=dSmJokMxTw@mail.gmail.com>
2013-05-27 16:36   ` Clemens Eisserer
2013-05-27 16:42 ` George Mitchell [this message]
2013-05-27 17:47 ` Sergei Trofimovich
2013-05-27 22:00   ` Szőts Ákos

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=51A38CE5.1010003@chinilu.com \
    --to=george@chinilu.com \
    --cc=linux-btrfs@vger.kernel.org \
    --cc=szotsaki@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).