[prev in list] [next in list] [prev in thread] [next in thread] 

List:       netatalk
Subject:    Re: [Netatalk-admins] Too many open files (probably an FD leak)
From:       Don Lee <netatalk () c ! icompute ! com>
Date:       2014-03-06 20:31:08
Message-ID: p06240803cf3e8b3e5c3f () [71 ! 39 ! 101 ! 61]
[Download RAW message or body]

[Attachment #2 (multipart/alternative)]


I'll try it, and report back. Thanks.

-dgl-

At 2:49 PM -0500 3/6/14, Arnaud Bergeron wrote:
This change fixes all my problems: 
<https://github.com/abergeron/Netatalk/commit/1f94c6c319cd1436e0873af0315560a8f23d1ef1 \
>https://github.com/abergeron/Netatalk/commit/1f94c6c319cd1436e0873af0315560a8f23d1ef1
> 

However, according to the developers it is not the right fix and will 
not be merged since it may cause other problems.  Therefore use with 
caution.



2014-03-06 12:41 GMT-05:00 Don Lee 
<<mailto:netatalk@c.icompute.com>netatalk@c.icompute.com>:

After digging around in the docs a bit, I learned that I am running with
Applecouble = v2 by virtue of running netatalk on a FS that does
not support native ea. It therefore makes sense that this is the
bug I have been hitting.

I have some trouble reading these threads in git/sourceforge because
it is not always clear chich fix/patch the comments are referring to.

Does anyone have a handle on a fix - even a suboptimal one - for this
problem?

I am hitting this problem, and the nasty part is that it runs the system
out of file table entries, and causes unrelated processes to fail.

I would be happy to stress test on a patch that someone believes to
be a fix.

Thnks,

-dgl-

So it appears I had some details wrong.  I've given a detailed 
account of all the facts that I know in a bug report, along with a 
debug trace of afpd while reproducing the problem.

If you want to resolve this, please refer to: 
<https://sourceforge.net/p/netatalk/bugs/554/>https://sourceforge.net/p/netatalk/bugs/554/




2014-03-01 18:13 GMT-05:00 Arnaud Bergeron 
<<mailto:abergeron@gmail.com>abergeron@gmail.com>:

I had it happen again now.  Since this is the weekend, I can do some 
debugging. 

In my afp.conf I have

cnid scheme = dbd

ea = ad

appledouble = v2


(I previously had ea = sys, but there are some serious stability 
issues with the ea implementation of ZFS on FreeBSD which lead to 
corrupted filesystems that hang the machine)

 From some debugging tests I made the fd that leaks is the mfd or rfd 
(both are the same).  The problem occurs when you click on an image 
file from a share in the finder while the window is in columns mode. 
This causes the little preview to appear to the right and makes apfd 
call ad_open_hf_v2() twice ending up with a ref count of 2 for 
mfd/rfd.  Then ad_close is only called once and the fd leaks.


I don't know if the bug is in the refcount of 2 or in the fact that 
ad_close() is only called once, while ad_open() is called twice.



------------------------------------------------------------------------------
Flow-based real-time traffic analytics software. Cisco certified tool.
Monitor traffic, SLAs, QoS, Medianet, WAAS etc. with NetFlow Analyzer
Customize your own dashboards, set traffic alerts and generate reports.
Network behavioral analysis & security monitoring. All-in-one tool.
<http://pubads.g.doubleclick.net/gampad/clk?id=126839071&iu=/4140/>http://pubads.g.doubleclick.net/gampad/clk?id=126839071&iu=/4140/ostg.clktrk


_______________________________________________
Netatalk-admins mailing list
<mailto:Netatalk-admins@lists.sourceforge.net>Netatalk-admins@lists.sourceforge.net
<https://lists.sourceforge.net/lists/listinfo/netatalk-admins>https://lists.sourceforge.net/lists/listinfo/netatalk-admins



[Attachment #5 (text/html)]

<!doctype html public "-//W3C//DTD W3 HTML//EN">
<html><head><style type="text/css"><!--
blockquote, dl, ul, ol, li { padding-top: 0 ; padding-bottom: 0 }
 --></style><title>Re: [Netatalk-admins] Too many open files (probably
an FD</title></head><body>
<div>I'll try it, and report back. Thanks.</div>
<div><br></div>
<div>-dgl-</div>
<div><br></div>
<div>At 2:49 PM -0500 3/6/14, Arnaud Bergeron wrote:</div>
<div>This change fixes all my problems: <a
href=
"https://github.com/abergeron/Netatalk/commit/1f94c6c319cd1436e0873af0315560a8f23d1ef1"><span
></span
>https://github.com/abergeron/Netatalk/commit/1f94c6c319cd1436e0873af<span
></span>0315560a8f23d1ef1</a><br>
</div>
<div>However, according to the developers it is not the right fix and
will not be merged since it may cause other problems.&nbsp; Therefore
use with caution.<br>
</div>
<div><br>
<br>
</div>
<div>2014-03-06 12:41 GMT-05:00 Don Lee &lt;<a
href="mailto:netatalk@c.icompute.com">netatalk@c.icompute.com</a>&gt;:<br
>
</div>
<blockquote>After digging around in the docs a bit, I learned that I
am running with</blockquote>
<blockquote>Applecouble = v2 by virtue of running netatalk on a FS
that does</blockquote>
<blockquote>not support native ea. It therefore makes sense that this
is the</blockquote>
<blockquote>bug I have been hitting.</blockquote>
<blockquote><br></blockquote>
<blockquote>I have some trouble reading these threads in
git/sourceforge because</blockquote>
<blockquote>it is not always clear chich fix/patch the comments are
referring to.</blockquote>
<blockquote><br></blockquote>
<blockquote>Does anyone have a handle on a fix - even a suboptimal one
- for this</blockquote>
<blockquote>problem?</blockquote>
<blockquote><br></blockquote>
<blockquote>I am hitting this problem, and the nasty part is that it
runs the system</blockquote>
<blockquote>out of file table entries, and causes unrelated processes
to fail.</blockquote>
<blockquote><br></blockquote>
<blockquote>I would be happy to stress test on a patch that someone
believes to</blockquote>
<blockquote>be a fix.</blockquote>
<blockquote><br></blockquote>
<blockquote>Thnks,</blockquote>
<blockquote><br></blockquote>
<blockquote>-dgl-</blockquote>
<blockquote><br></blockquote>
<blockquote>So it appears I had some details wrong.&nbsp; I've given a
detailed account of all the facts that I know in a bug report, along
with a debug trace of afpd while reproducing the problem.<br>
</blockquote>
<blockquote>If you want to resolve this, please refer to: <a
href="https://sourceforge.net/p/netatalk/bugs/554/"
>https://sourceforge.net/p/netatalk/bugs/554/</a><br>
</blockquote>
<blockquote><br>
<br>
</blockquote>
<blockquote>2014-03-01 18:13 GMT-05:00 Arnaud Bergeron &lt;<a
href="mailto:abergeron@gmail.com">abergeron@gmail.com</a>&gt;:<br>
<blockquote>I had it happen again now.&nbsp; Since this is the
weekend, I can do some debugging.&nbsp;<br>
</blockquote>
<blockquote>In my afp.conf I have<br>
</blockquote>
<blockquote>cnid scheme = dbd<br>
</blockquote>
<blockquote>ea = ad<br>
</blockquote>
<blockquote>appledouble = v2<br>
</blockquote>
<blockquote><br></blockquote>
<blockquote>(I previously had ea = sys, but there are some serious
stability issues with the ea implementation of ZFS on FreeBSD which
lead to corrupted filesystems that hang the machine)<br>
</blockquote>
<blockquote>From some debugging tests I made the fd that leaks is the
mfd or rfd (both are the same).&nbsp; The problem occurs when you
click on an image file from a share in the finder while the window is
in columns mode.&nbsp; This causes the little preview to appear to the
right and makes apfd call ad_open_hf_v2() twice ending up with a ref
count of 2 for mfd/rfd.&nbsp; Then ad_close is only called once and
the fd leaks.<br>
</blockquote>
<blockquote><br></blockquote>
<blockquote>I don't know if the bug is in the refcount of 2 or in the
fact that ad_close() is only called once, while ad_open() is called
twice.<br>
</blockquote>
</blockquote>
<blockquote><br></blockquote>
<blockquote><br>
---------------------------------------------------------------------<span
></span>---------<br>
Flow-based real-time traffic analytics software. Cisco certified
tool.<br>
Monitor traffic, SLAs, QoS, Medianet, WAAS etc. with NetFlow
Analyzer<br>
Customize your own dashboards, set traffic alerts and generate
reports.<br>
Network behavioral analysis &amp; security monitoring. All-in-one
tool.<br>
<a
href=
"http://pubads.g.doubleclick.net/gampad/clk?id=126839071&amp;iu=/4140/"
>http://pubads.g.doubleclick.net/gampad/clk?id=126839071&amp;iu=/4140/</a
>ostg.clktrk<br>
<br>
_______________________________________________<br>
Netatalk-admins mailing list<br>
<a
href="mailto:Netatalk-admins@lists.sourceforge.net"
>Netatalk-admins@lists.sourceforge.net</a><br>
<a
href="https://lists.sourceforge.net/lists/listinfo/netatalk-admins"
>https://lists.sourceforge.net/lists/listinfo/netatalk-admins</a></blockquote
>
<div><br></div>
</body>
</html>

------------------------------------------------------------------------------
Subversion Kills Productivity. Get off Subversion & Make the Move to Perforce.
With Perforce, you get hassle-free workflows. Merge that actually works. 
Faster operations. Version large binaries.  Built-in WAN optimization and the
freedom to use Git, Perforce or both. Make the move to Perforce.
http://pubads.g.doubleclick.net/gampad/clk?id=122218951&iu=/4140/ostg.clktrk

_______________________________________________
Netatalk-admins mailing list
Netatalk-admins@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/netatalk-admins


[prev in list] [next in list] [prev in thread] [next in thread] 

Configure | About | News | Add a list | Sponsored by KoreLogic