<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
On 27/09/11 12:43, Aurélien Lemaire wrote:
<blockquote cite="mid:4E81A8C4.7030906@smile.fr" type="cite">
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
<br>
Le 16/09/2011 08:55, Jorge Nerín a écrit :
<blockquote
cite="mid:CAF-dNMaKvvxnHCiTR2-jVQyRC9s7Bi4J81hUTh7x1RW9-___Jw@mail.gmail.com"
type="cite"><br>
<div class="gmail_quote">On Thu, Sep 15, 2011 at 11:29, Aurélien
Lemaire <span dir="ltr"><<a moz-do-not-send="true"
href="mailto:aurelien.lemaire@smile.fr">aurelien.lemaire@smile.fr</a>></span>
wrote:<br>
<blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt
0.8ex; border-left: 1px solid rgb(204, 204, 204);
padding-left: 1ex;">
<div bgcolor="#ffffff" text="#000000"> Good day folks,<br>
<br>
First of all, varnish is an outstanding piece of software
that my company and i are addicted to. So thanks to all
the coders.<br>
<br>
Here is my problem :<br>
I allocated varnish 1G of RAM on a website that can have
more than 2 Go of possible cacheable objects . Not to
worry though as any proxy-cache system should smartly nuke
old objects to make place to new one to live peacefully
within its allocated RAM. And that's where Varnish behave
unexpectedly : each time it need to nuke SOME objects : it
nukes ALMOST ALL of them (often ~80% of my 35k objects)
which is quite aggressive ; thus i lost almost all my
cache....IRK !<br>
<br>
3 Munin graphs attached to see the problem clearly : big
drop each time a nuking happens.<br>
<br>
To make sure my pbr is about varnish nuking system : i
increased from 1G to 3G(more than the max possible 2G
cacheable objects) on another varnish of this platefom
(this website is delivered by multiple front/varnish
server all stricly similar and independant) and this issue
disappeared (no more nuking : no lost of ~80%of my
objects)<br>
<br>
Here is my env : <br>
Debian 5.0.8 64 bits on 2.6.32-5-openvz-amd64 kernel<br>
Varnish 2.1.3 SVN 5049:5055(debian package 2.1.3-8) <br>
200 varnish 's worker threads running constantly (no issue
on workers)<br>
30req/s average with 60/s in peak<br>
<br>
Daemon run as such :<br>
/usr/sbin/varnishd -P /var/run/varnishd.pid -a :80 -T
localhost:6082 -S /etc/varnish/secret -f
/etc/varnish/serverx.vcl -w 100,1024 -s
file,/var/lib/varnish/serverx/varnish_storage.bin,3G<br>
<br>
Here a quick varnishstat -1 :<br>
<br>
Is it normal varnish behaviour ? sounds like a bug to me.<br>
Am i missing some tuning (lru_interval) to soften the
nuking algo ?<br>
Do you need more info ?<br>
helps appreciated here ;-)<br>
<br>
Regards, Aurelien Lemaire<br>
</div>
<br>
</blockquote>
</div>
<br>
It could be someone downloading a large file (like a ~700Mb iso
file) and varnish nuking objects to make room for this file
(even if its configured to not cache it).<br>
<br>
Try to get a varnishlog trace of the moment the nuking begins.<br
clear="all">
<br>
-- <br>
Jorge Nerín<br>
<<a moz-do-not-send="true" href="mailto:jnerin@gmail.com">jnerin@gmail.com</a>><br>
</blockquote>
This website does not content any files bigger than a couple of
Mb. <br>
<br>
Will see what i can do to get the varnishlog during the nuking.<br>
<br>
--<br>
aurelien lemaire<br>
</blockquote>
<br>
<font face="Arial">Providing a reduced-to-the-minimum varnishlog on
this stressed prod env won't be easy for me moreover i doubt that
nuking activity is reported by varnishlog . So meanwhile </font> :<br>
<br>
<font face="Arial">My "agressive nuking" problem seems to be known
and often-encountered :<br>
- </font><font face="Arial">
<a class="moz-txt-link-freetext" href="http://www.gossamer-threads.com/lists/varnish/dev/9631">http://www.gossamer-threads.com/lists/varnish/dev/9631</a><br>
- </font><font face="Arial" color="black"><font color="black"><a class="moz-txt-link-freetext" href="http://www.mail-archive.com/varnish-misc@projects.linpro.no/msg03684.html">http://www.mail-archive.com/varnish-misc@projects.linpro.no/msg03684.html</a><br>
- Quite some thread without any outcome found by google search
where all the people just fix it by increasign the the "-s file"
size which only postpone the problem <br>
<br>
On this first link </font></font><font face="Arial">Artur ( a
varnish dev
<a class="moz-txt-link-freetext" href="http://www.gossamer-threads.com/lists/engine?user=137;list=varnish">http://www.gossamer-threads.com/lists/engine?user=137;list=varnish</a>
) shortly admit "</font><font face="Arial" color="black"><font
color="black"><i>... There are problems with the fragmentation
of the store...</i>" </font></font><font face="Arial">on the
1st of april 2009</font><font face="Arial" color="black"><font
color="black">.</font></font><br>
On the second link there is all the varnishlog and backtrace infos
needed to open the bug report to get it fixed<br>
<font color="black"><font face="Arial"><br>
</font></font><font face="Arial">Thus is it possible to open this
bug report officially as "to be fixed" ?</font><br>
<br>
At the same time i'll try to swith to "-s malloc " as suggested by
Artur and keep you all informed.<br>
<br>
<font face="Arial">Cheers, </font><br>
<font face="Arial">--</font><br>
<font face="Arial">Aurélien Lemaire</font><br>
<br>
</body>
</html>