Seg fault with Image::Magick on FreeBSD
As soon as I preload Image::Magick under mod_perl2 on FreeBSD (5.3, 5.4 and 6.0) I get a "Segmentation fault (core dumped)" when I start Apache. Without preloading it works fine. What's the best place to report this bug ?
Re: Seg fault with Image::Magick on FreeBSD
> I'm going to have to eat part/all of my words here as there is no port for > p5-Image-Magick. There is a port ImageMagick which also contains the perl module Image::Magick. I've mailed the maintainer. Not sure though if he would be able to fix it as he might not be into perl.
Re: Seg fault with Image::Magick on FreeBSD
> Somewhere, it uses $& or one of those operators that change the regex engine of > perl. I don't know if its been fixed in 5.8.7 or maint. I'm running 5.8.7 and it still has this problem. I'm currently working around it by not preloading Image::Magick and using "require" to load Image::Magick on places where it's needed. It works but it's ugly Len
Re: ImageMagick
> trunks# apachectl start > Segmentation fault (core dumped) > > Of course, the FreeBSD pkg builds are without debugging. > I'll give that a whirl shortly. Thanks for that Philip. I know perl but this is way beyond my reach.
Out of memory ?!
Hello all, Once in a while I get "Out of memory" errors on our mod_perl servers. They look like this: Out of memory during request for 77000 bytes, total sbrk() is 14952448 bytes! [Mon Nov 7 00:08:01 2005] thaw.al): Callback called exit at ../../lib/Storable.pm (autosplit into ../../lib/auto/Storable/tha w.al) line 363. At last Apache crashes completely. It's not related to a specific application: It happens on our WebGUI (= cms) nodes but also on our (relatively simple) custom applications that run on different boxes. All perl stuff is running under ModPerl::Registry. We're using FreeBSD 5.3 / apache-2.0.50_3 / mod_perl2-1.99r16 / perl-5.8.5 all installed using the ports mechanism. I'm using MaxRequestPerChild 600 to limit the http size growth. I've tried to restart apache every night but I still get that "Out of Memory" errors. Anyone familiar with this problem ? Len
Re: Out of memory ?!
> Are they always in Storable::thaw? What are you thawing? No, it happens on any module: Out of memory during request for 55504 bytes, total sbrk() is 26630144 bytes! [Fri Dec 2 11:24:11 2005] TagFilter.pm: Callback called exit at /knvb/WebGUI/li b/HTML/TagFilter.pm line 402. [Fri Dec 2 11:24:11 2005] Carp.pm: Callback called exit at /usr/local/lib/perl5 /5.8.5/CGI/Carp.pm line 314. > What code is running on both of these? A wide range of perl modules, like HTML::Template, XML::Simple, etc, etc. I'm going to upgrade to latest mod_perl2 and start using Apache2::SizeLimit and Apache2::Resource. Hope that will fix it.
Why is my apache parent process growing...
Happy 2006 everyone, I'm running Apache/2.0.54, mod_perl/2.0.1, Perl/v5.8.7 on FreeBSD 6.0. Right after starting apache (with preloading the needed perl modules) the httpd root process is about 40MB. After several days it has grown to over 100MB. I'm using Apache2::SizeLimit to kill of httpd childs that get too large. I'm using this configuration: $Apache2::SizeLimit::MAX_PROCESS_SIZE = 10; $Apache2::SizeLimit::MAX_UNSHARED_SIZE = 25000; The problem here is that when the Apache parent process gets 10 KB big, the spawned childs also are that size, which means that a spawned child is killed immediately after it served its first request. The only way to decrease the parent size is to stop and start apache. My understanding is that the parent process is not supposed to serve any requests itself. It's entire job is to manage the children. So, how come it is growing, and more importantly, how can I control it ? BTW: I have this problem on all our FreeBSD boxes running different applications. Is FreeBSD / mod_perl just a bad combo ? Len Kranendonk
Re: Why is my apache parent process growing...
Do you have any globals or other shared information that grows very large? Do you read in any large data sets either from a database or a file into the application all at once? We're using a global %session hash, which is undef 'fed when the session is closed. I'm not reading in large data sets, but even if I did that should grow the child process, not the parent process, right ? The application I'm talking about is WebGUI, a CMS written in perl. I've mailed the WebGUI list earlier today. It seems that the people running on Linux don't have the problem that's causing me pain. That's why I start to believe that it's a FreeBSD/mod_perl issue. The problem is not limited to WebGUI, I also have it on two boxes that serve football results. The only thing that's running there is SOAP::Lite, HTML::Template, XML::Simple and Cache::FileCache. After two months that http root process has grown from 20MB to 185MB. Len
Re: Why is my apache parent process growing...
How are you measuring this? top, or ps aux and watch the RSS column (the real memory (resident set) size of the process (in 1024 byte units)). A sample top output: PID USERNAME PRI NICE SIZERES STATETIME WCPUCPU COMMAND 87075 www40 178M 174M accept 0:32 3.08% 3.08% httpd 86839 www40 175M 171M accept 1:53 0.39% 0.39% httpd 86827 www40 188M 184M accept 2:12 0.00% 0.00% httpd 86836 www40 187M 183M accept 2:06 0.00% 0.00% httpd 86830 www40 172M 168M accept 2:02 0.00% 0.00% httpd 86831 www40 177M 173M accept 2:02 0.00% 0.00% httpd 86858 www40 189M 185M accept 1:58 0.00% 0.00% httpd 87070 www40 170M 167M accept 0:49 0.00% 0.00% httpd 87073 www40 187M 184M accept 0:38 0.00% 0.00% httpd 87074 www40 174M 170M accept 0:29 0.00% 0.00% httpd 69201 root 960 157M 154M select 0:13 0.00% 0.00% httpd The initial size of the httpd root process was 46 MB. That's correct. Are you using Cache::FastMmap, or anything else that does memory-mapping of files? That might yield an apparent larger size if the parent process is one of the ones that has the mmap'ed file open. No, I'm not using Cache::FastMmap.
Re: Why is my apache parent process growing...
Are you using IPC::SharedCache for your HTML::Template caching? No, I'm using the "file_cache=>1" option, which uses Storable (I believe).
Re: Why is my apache parent process growing...
This doesn't really help you, but I don't notice the parent growing on any of my FreeBSD boxes doing a variety of things. So, most likely I'm doing something that you don't do Are you using ModPerl::Registry, because I still do.
Found it !!! Why is my apache parent process growing...
I found it !!! Issueing an "apachectl restart" (or kill -HUP) make the apache parent process size grow ! This is the case on all of our FreeBSD boxes. I can't reproduce it on Linux. Take a look at this: apachectl start # ps aux|grep httpd | grep root root 70820 59.0 5.6 46408 43552 ?? Ss4:03PM 0:04.40 /usr/local/sbin/httpd -k start So: 43MB right after starting Now send it a SIGHUP: # kill -HUP 70820 # ps aux| grep httpd | grep root root 70820 17.4 7.7 62440 59564 ?? Rs4:03PM 0:08.48 /usr/local/sbin/httpd -k start Wow: the HTTP root process has grown from 43 to 59MB just because I sent it a SIGHUP. On my test box it will take about 50 "apachectl restart" commands, but then apache crashes with a "Out of memory during request" error. Philip, as you're using FreeBSD too, could you confirm this ? I'm using a log rotation script which send a SIGHUP to apache every night, so that was actually causing the apache root process to grow. Thanks anyone for helping me to find this problem ! Len
Why is my apache parent process growing...
Hello all, I'm running Apache/2.0.54, mod_perl/2.0.1, Perl/v5.8.7 on FreeBSD 6.0. Right after starting apache (with preloading the needed perl modules) the httpd root process is about 40MB. After several days it has grown to over 100MB. I'm using Apache2::SizeLimit to kill of httpd childs that get too large. I'm using this configuration: $Apache2::SizeLimit::MAX_PROCESS_SIZE = 10; $Apache2::SizeLimit::MAX_UNSHARED_SIZE = 25000; The problem here is that when the Apache parent process gets 10 KB big, the spawned childs also are that size, which means that a spawned child is killed immediately after it served its first request. The only way to decrease the parent size is to stop and start apache. My understanding is that the parent process is not supposed to serve any requests itself. It's entire job is to manage the children. So, how come it is growing, and more importantly, how can I control it ? BTW: I have this problem on all our FreeBSD boxes running different applications. Is FreeBSD / mod_perl just a bad combo ? Len Kranendonk
What handler am I ?
Hello all,
Suppose I have this:
PerlModule My::Module
PerlChildInitHandler My::Module
PerlOutputFilterHandler My::Module
In the sub handler {} , is there a way to
determinate what type of handler I am supposed to be ?
I'm running mod_perl2 for that
matters.
Len
Re: What handler am I ?
Exactly what I was look for ! Thanks !
- Original Message -
From: "Geoffrey Young" <[EMAIL PROTECTED]>
To: "Len Kranendonk" <[EMAIL PROTECTED]>
Cc:
Sent: Wednesday, February 01, 2006 3:32 PM
Subject: Re: What handler am I ?
Len Kranendonk wrote:
Hello all,
Suppose I have this:
PerlModule My::Module
PerlChildInitHandler My::Module
PerlOutputFilterHandler My::Module
In the sub handler {} , is there a way to determinate what type of
handler I am supposed to be ?
yes, current_callback()
I'm running mod_perl2 for that matters.
yes, it does :)
for mp2 it's
use ModPerl::Util;
my $callback = ModPerl::Util::current_callback();
for mp1 it's simply
my $callback = $r->current_callback();
HTH
--Geoff
Re: What handler am I ?
my $callback = ModPerl::Util::current_callback(); For a PerlOutputFilterHandler the current_callback() function returns "PerlResponseHandler". That sounds like a bug, or am I doing something wrong ? Server: Apache/2.0.54 (FreeBSD 6.0) mod_ssl/2.0.54 OpenSSL/0.9.7g mod_apreq2-20050712/2.1.3-dev mod_perl/2.0.1 Perl/v5.8.7
Re: clusters
And also this: "NDB is an in-memory storage engine" No longer true: As of MySQL 5.1.6 it's possible to store tables on disk rather than in memory. See http://dev.mysql.com/doc/refman/5.1/en/mysql-cluster-disk-data.html But, I wouldn't recommend using MySQL cluster to power your high traffic website. I've tested it but the throughput is at least 10 - 100x worse in compare to a good old MySQL master / slave setup on the same hardware. Esspecially the writes are slow. SQL Relay offers load balancing functionality, without having to change your existing code: http://sqlrelay.sourceforge.net/ Len
