[Bioperl-l] Fwd: [Gmod-gbrowse] sleeping mysql processes w/Bio::DB::SeqFeature::Store

Jason Stajich jason.stajich at gmail.com
Wed Jul 10 23:16:52 UTC 2013


Another reason to make sure this module is its own package so Lincoln can release it more rapidly than BioPerl. Hope new branch can fix that. 

Sent from my iPhone-please excuse typos

--
Jason Stajich

Begin forwarded message:

> From: Lincoln Stein <lincoln.stein at gmail.com>
> Date: July 10, 2013, 7:58:05 PDT
> To: Martin Mokrejs <mmokrejs at fold.natur.cuni.cz>
> Cc: Jim Hu <jimhu at tamu.edu>, GMOD GBrowse List <gmod-gbrowse at lists.sourceforge.net>
> Subject: Re: [Gmod-gbrowse] sleeping mysql processes w/Bio::DB::SeqFeature::Store
> 
> I have just uploaded GBrowse version 2.55 to CPAN. It contains a bioperl
> patch that fixes slow loading of multiple GFF3 files, and changes the
> timeout values for mod_fastcgi and mod_fcgid. I have also noticed that
> mod_fastcgi now seems to be faster than mod_fcgid when the timeout issue is
> fixed.
> 
> I don't think that the FastCGI issues are related to the sleeping mysql
> process issue, actually, because this was first reported in the context of
> running GBrowse in CGI mode. I am starting work on CGI mode now.
> 
> Lincoln
> 
> 
> On Tue, Jul 9, 2013 at 6:49 PM, Martin Mokrejs
> <mmokrejs at fold.natur.cuni.cz>wrote:
> 
>> Hi,
>> 
>> Lincoln Stein wrote:
>>> Hi Folks,
>>> 
>>> With respect to the FastCGI errors, I have tracked the problem down to
>> the following sequence of events:
>>> 
>>> 1. The FastCGI executive module launches a fresh gbrowse instance.
>>> 2. GBrowse tries to load its default database into memory.
>>> 3. If the default database takes more than 3s to load, then FastCGI
>> times it out.
>>> 4. FastCGI launches a new instance of GBrowse.
>>> 5. GBrowse tries to load its default database into memory.
>>> 6. FastCGI times the new instance out.
>>> 7. Repeat 4-6 indefinitely.
>>> 
>>> Note that this only happens for databases that are slow to load,
>> typically in-memory databases. For example, the full tutorial database
>> takes ~8s to load on my machine (unreasonably slow for reasons that are
>> unclear to me: I am going to start debugging bioperl next). Once the
>> database is loaded, however, all subsequent accesses are fast.
>>> 
>>> Here's what to do to fix the problem:
>>> 
>>>  * for mod_fcgid, add the following directive to
>> /etc/apache2/conf.d/gbrowse2 (in the same section as *FcgidIOTimeout*)
>>>      o   *FcgidConnectTimeout 30*
>>>  * for mod_fastcgi, change *FastCGIConfig* to
>>>      o *FastCGIConfig *-startDelay 30 -appConnTimeout 30 -idle-timeout
>> 600 -maxClassProcesses 20  -initial-env GBROWSE_CONF=/etc/gbrowse2
>> 
>> Hmm, I would call this a workaround instead. Why isn't there something
>> like a lock
>> in Gbrowse so that multiple connections to mysql wouldn't be established?
>> This
>> will definitely trick again some user or sysadmin. At least if gbrowse
>> would test
>> for a timeout value and complain and exit if it would be too short.
>> 
>> In my eyes gbrowse is full of such tricks and reading this lists only
>> ensures me that
>> the code should be more careful about what user has configured, what is
>> and what is not
>> available, and ... provide helpful error messages. I don't have the time
>> to contribute
>> some code in this regard but anything leading to perl unitialized values
>> should be
>> fixed. At least, I could fish some emails from the archives and emphasize
>> the worst
>> examples. Or the other way around, jsut take a working setup and start to
>> screw file/dir
>> permission, place bad filenames/dbnames/dirnames into config, set screw
>> variable names.
>> It will all generate all kinds of funny messages to apache logs and that
>> should fairly
>> doable to put a couple of roadblocks along the way.
>> 
>> This would only be helpful to everybody asking here on this list for help
>> with some weird
>> messages in apache logs.
>> 
>> Just my 2cents, ;-)
>> Martin
> 
> 
> 
> -- 
> Lincoln D. Stein
> Director, Informatics and Biocomputing Platform
> Ontario Institute for Cancer Research
> 101 College St., Suite 800
> Toronto, ON, Canada M5G0A3
> 416 673-8514
> Assistant: Renata Musa <Renata.Musa at oicr.on.ca>
> ------------------------------------------------------------------------------
> See everything from the browser to the database with AppDynamics
> Get end-to-end visibility with application monitoring from AppDynamics
> Isolate bottlenecks and diagnose root cause in seconds.
> Start your free trial of AppDynamics Pro today!
> http://pubads.g.doubleclick.net/gampad/clk?id=48808831&iu=/4140/ostg.clktrk
> _______________________________________________
> Gmod-gbrowse mailing list
> Gmod-gbrowse at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/gmod-gbrowse




More information about the Bioperl-l mailing list