- dancer-users - lists.preshweb.co.uk

Re: [Dancer-users] factory recommendation for web server
by Puneet Kishor 05 Jan '11

05 Jan '11

Thanks Deepak. A few follow-up questions (others, please feel free to jump in with your wisdom) -- Deepak Gulati wrote: > I run Dancer under Starman and use nginx as a reverse caching proxy. I > also let nginx serve most static files (public/images for example) and > configure it to gzip the response, add expires headers etc. IMHO > Dancer should do what it does best and nginx should do all the other > HTTP related heavy-lifting. Well, Dancer is not a web server. It still needs a web server. When you say that "Dancer should do what it does best and nginx should do all other" stuff, I am assuming what you are trying to say is that Nginx should server all static files, and <Starman> should server all non-static files. (I put <Starman> in <> because one could replace it with another web server). My question is thus -- why need two different servers? Can't Nginx do everything? After all, Dancer doesn't serve and content. Dancer creates the content, or routes the web server to the correct content. It is the web server that serves the content. Doesn't having one web server serve static content and the other serve Dancer generated content make your web application more complicated? I have only one web server right now -- just Apache. If I replace Apache, I want to replace it with something lighter weight, but equally capable, AND, only one program, not multiple programs. > Starman is probably an overkill for my > setup. Why is Starman a probable overkill? Does it consume too many resources? > > I run the dancer app using daemontools and the following run script: > > #!/bin/sh > > export PERL5LIB='/home/ec2-user/TwitterToys/lib' > exec 2>&1 \ > /usr/local/bin/plackup -s Starman -l 127.0.0.1:5001 -E deployment > --workers=10 -a /home/ec2-user/TwitterToys/bin/app.pl One more question -- Since you are running Starman on port 5001 (as shown above), what would you do for other apps? Would you have to start Starman on a different port for every app that you want to serve? This would mean keeping track of different Starman instances and the ports they were running on. > > Here is a snippet from my nginx config. Hopefully I'll be able to get > a detailed step by step write-up out by this weekend. > > http { > gzip on; > gzip_min_length 1024; > > proxy_cache_path /var/cache/nginx levels=1:2 keys_zone=twitter:8m > max_size=64m inactive=60m; > proxy_temp_path /tmp; > proxy_cache_key "$scheme://$host$request_uri"; > proxy_cache_valid 200 60m; > > server { > server_name stupidtwitterstats.com; > listen 80; > > location / { > proxy_cache twitter; > set $do_not_cache 0; > if ( $request_uri ~ "^/test$" ) { > set $do_not_cache 1; > } > proxy_no_cache $do_not_cache; > proxy_pass http://127.0.0.1:5001; > proxy_redirect http://127.0.0.1:5001/ http://$host/; > expires 1h; > } > > location /images/ { > alias /home/ec2-user/TwitterToys/public/images/; > expires 30d; > access_log off; > } > } > > On Tue, Jan 4, 2011 at 10:12 PM, Puneet Kishor<punk.kish(a)gmail.com> wrote: >> I know there are more than one ways to do it, but I have stuck with Apache >> for three reasons -- >> >> - historically Apache has been the front-runner although it may not be >> anymore; >> >> - at least one of the programs I use utilizes CGI, and I am only familiar >> with Apache's CGI capabilities; and >> >> - Apache comes installed by default on my computer, my server, and on most >> all shared web hosting services. >> >> That said, I do have control of my own computer and my server (although no >> control of the shared service for at least one site). What is, if any, the >> recommended combination of web servers/proxy servers, etc. that are optimal >> for Dancer deployment? >> >> The deployment guide does give information on many variations, but, if you, >> the master dancers, could choose any, what would you choose? >> >> I see a lot of mention of Nginx and Starman, and am considering >> experimenting with those. >> >> I want speed, ease of installation and configuration, detailed error (and >> access) logging, and extraction of maximum efficiency with Dancer. >> >> >> >> -- Puneet Kishor http://punkish.org Carbon Model http://carbonmodel.org Charter Member, Open Source Geospatial Foundation http://www.osgeo.org Science Fellow http://creativecommons.org/about/people/fellows#puneetkishor Nelson Institute, UW-Madison http://www.nelson.wisc.edu --------------------------------------------------------------------------- Assertions are politics; backing up assertions with evidence is science ===========================================================================

3 2

Fwd: factory recommendation for web server
by Deepak Gulati 05 Jan '11

05 Jan '11

Adding dancer-users back to the thread... Inline... On Wed, Jan 5, 2011 at 1:38 AM, Puneet Kishor <punk.kish(a)gmail.com> wrote: > Thanks Deepak. A few follow-up questions (others, please feel free to jump > in with your wisdom) -- > > > Deepak Gulati wrote: >> >> I run Dancer under Starman and use nginx as a reverse caching proxy. I >> also let nginx serve most static files (public/images for example) and >> configure it to gzip the response, add expires headers etc. IMHO >> Dancer should do what it does best and nginx should do all the other >> HTTP related heavy-lifting. > > Well, Dancer is not a web server. It still needs a web server. When you say > that "Dancer should do what it does best and nginx should do all other" > stuff, I am assuming what you are trying to say is that Nginx should server > all static files, and <Starman> should server all non-static files. (I put > <Starman> in <> because one could replace it with another web server). Indeed. Dancer should do dynamic content. And although there are quite a few capable web-servers that I can use today with plack and run Dancer 'standalone' without putting a web-server in front, I chose this setup because: - I get a lot of things for free with Nginx like caching. - I also get to use the knowledge of the nginx community for common chores like redirecting www.xyz.com to xyz.com. - I get to scale vertically by running multiple dancer instances on different ports or horizontally where one Nginx instance front-ends multiple dancer instances running on different machine - In my setup, I could have done without Starman and run as FCGI but I just found it convenient that I could troubleshoot issues (e.g. to check if my templates got updated while nginx is pushing out cached content) by directly connecting to the Dancer instance on port 5000 through my browser. > My question is thus -- why need two different servers? Can't Nginx do > everything? After all, Dancer doesn't serve and content. Dancer creates the > content, or routes the web server to the correct content. It is the web > server that serves the content. Doesn't having one web server serve static > content and the other serve Dancer generated content make your web > application more complicated? > > I have only one web server right now -- just Apache. If I replace Apache, I > want to replace it with something lighter weight, but equally capable, AND, > only one program, not multiple programs. > >> Starman is probably an overkill for my >> setup. > > Why is Starman a probable overkill? Does it consume too many resources? - It's lightweight enough but I could've just used FCGI since Nginx does all the HTTP front-ending work for me. That said, I too come from a Apache, mod_perl background so these are early days yet. > >> >> I run the dancer app using daemontools and the following run script: >> >> #!/bin/sh >> >> export PERL5LIB='/home/ec2-user/TwitterToys/lib' >> exec 2>&1 \ >> /usr/local/bin/plackup -s Starman -l 127.0.0.1:5001 -E deployment >> --workers=10 -a /home/ec2-user/TwitterToys/bin/app.pl > > One more question -- Since you are running Starman on port 5001 (as shown > above), what would you do for other apps? Would you have to start Starman on > a different port for every app that you want to serve? This would mean > keeping track of different Starman instances and the ports they were running > on. I haven't given this a thought since I was deploying just one app. But yes, I can see it becoming unmanageable quickly. >> >> Here is a snippet from my nginx config. Hopefully I'll be able to get >> a detailed step by step write-up out by this weekend. >> >> http { >> gzip on; >> gzip_min_length 1024; >> >> proxy_cache_path /var/cache/nginx levels=1:2 keys_zone=twitter:8m >> max_size=64m inactive=60m; >> proxy_temp_path /tmp; >> proxy_cache_key "$scheme://$host$request_uri"; >> proxy_cache_valid 200 60m; >> >> server { >> server_name stupidtwitterstats.com; >> listen 80; >> >> location / { >> proxy_cache twitter; >> set $do_not_cache 0; >> if ( $request_uri ~ "^/test$" ) { >> set $do_not_cache 1; >> } >> proxy_no_cache $do_not_cache; >> proxy_pass http://127.0.0.1:5001; >> proxy_redirect http://127.0.0.1:5001/ http://$host/; >> expires 1h; >> } >> >> location /images/ { >> alias /home/ec2-user/TwitterToys/public/images/; >> expires 30d; >> access_log off; >> } >> } >> >> On Tue, Jan 4, 2011 at 10:12 PM, Puneet Kishor<punk.kish(a)gmail.com> >> wrote: >>> >>> I know there are more than one ways to do it, but I have stuck with >>> Apache >>> for three reasons -- >>> >>> - historically Apache has been the front-runner although it may not be >>> anymore; >>> >>> - at least one of the programs I use utilizes CGI, and I am only familiar >>> with Apache's CGI capabilities; and >>> >>> - Apache comes installed by default on my computer, my server, and on >>> most >>> all shared web hosting services. >>> >>> That said, I do have control of my own computer and my server (although >>> no >>> control of the shared service for at least one site). What is, if any, >>> the >>> recommended combination of web servers/proxy servers, etc. that are >>> optimal >>> for Dancer deployment? >>> >>> The deployment guide does give information on many variations, but, if >>> you, >>> the master dancers, could choose any, what would you choose? >>> >>> I see a lot of mention of Nginx and Starman, and am considering >>> experimenting with those. >>> >>> I want speed, ease of installation and configuration, detailed error (and >>> access) logging, and extraction of maximum efficiency with Dancer. >>> >>> >>> >>> > > > > -- > Puneet Kishor http://punkish.org > Carbon Model http://carbonmodel.org > Charter Member, Open Source Geospatial Foundation http://www.osgeo.org > Science Fellow http://creativecommons.org/about/people/fellows#puneetkishor > Nelson Institute, UW-Madison http://www.nelson.wisc.edu > --------------------------------------------------------------------------- > Assertions are politics; backing up assertions with evidence is science > =========================================================================== >

1 0

I launched my first dancer app
by Deepak Gulati 04 Jan '11

04 Jan '11

Hi Folks, Just wanted to drop a word about my first dancer app (Stupid Twitter Stats): http://stupidtwitterstats.com/ Although I've been dabbling with Dancer since Apr 2010, I never quite got down to writing (or launching) something worthwhile. This past week I sat down and decided to see this little idea through. I thoroughly enjoyed working with Dancer and will definitely be using it more in the coming days. I am deployed on EC2 with nginx as a reverse proxy to Dancer running on plack/Starman. I'll shortly write down my experience of setting the production app along with my nginx config file and drop a note here. Thanks everyone for making it possible! Cheers, Deepak

4 4

detailed logs, not
by Puneet Kishor 04 Jan '11

04 Jan '11

This is befuddling. My config files has the following incantation config.yml --------------------- logger : "file" development.yml --------------------- log : "debug" warnings : 1 show_errors : 1 Yet, no detailed warnings or errors are shown anywhere. If I add a specific debug statement, then that shows up in the "development.log." A few warnings do show up from time to time in the Apache log. I can't place a pattern on it -- some kinds of errors show up in the Apache log, some, at times, are returned in the http response and are visible in Firebug, and some don't show up at all. For example, I had the following simple open my $fh, ">", $file or die "$file: $!\n"; There was a permission problem, and the `open` command was die-ing, but the error was not visible anywhere. I changed the line to open my $fh, ">", $file or debug "$file: $!\n"; and I got the error "permission denied" in my development.log. The above is a simple instance. I am actually using PDL (Perl Data Language) for munging through large datasets to display in the web application. None of my PDL errors are showing up in any log. The only alternative for me is to very tediously insert the `debug` command in different places until I determine what command is breaking. Any suggestions on how I could get detailed warnings and errors consistently in one place? -- Puneet Kishor http://punkish.org Carbon Model http://carbonmodel.org Charter Member, Open Source Geospatial Foundation http://www.osgeo.org Science Fellow http://creativecommons.org/about/people/fellows#puneetkishor Nelson Institute, UW-Madison http://www.nelson.wisc.edu --------------------------------------------------------------------------- Assertions are politics; backing up assertions with evidence is science ===========================================================================

2 5

factory recommendation for web server
by Puneet Kishor 04 Jan '11

04 Jan '11

I know there are more than one ways to do it, but I have stuck with Apache for three reasons -- - historically Apache has been the front-runner although it may not be anymore; - at least one of the programs I use utilizes CGI, and I am only familiar with Apache's CGI capabilities; and - Apache comes installed by default on my computer, my server, and on most all shared web hosting services. That said, I do have control of my own computer and my server (although no control of the shared service for at least one site). What is, if any, the recommended combination of web servers/proxy servers, etc. that are optimal for Dancer deployment? The deployment guide does give information on many variations, but, if you, the master dancers, could choose any, what would you choose? I see a lot of mention of Nginx and Starman, and am considering experimenting with those. I want speed, ease of installation and configuration, detailed error (and access) logging, and extraction of maximum efficiency with Dancer. -- Puneet Kishor http://punkish.org Carbon Model http://carbonmodel.org Charter Member, Open Source Geospatial Foundation http://www.osgeo.org Science Fellow http://creativecommons.org/about/people/fellows#puneetkishor Nelson Institute, UW-Madison http://www.nelson.wisc.edu --------------------------------------------------------------------------- Assertions are politics; backing up assertions with evidence is science ===========================================================================

1 0

config.yml
by Todd Bruner 04 Jan '11

04 Jan '11

Could some one point me to at what I'm doing wrong in the config.yml? I'm following the tutorial and the advent stuff and trying to enable plack_middlewares. Here's my config.yml layout: "main" charset: "UTF-8" template: "template_toolkit" engines: template_toolkit: encoding: 'utf8' start_tag: '[%' end_tag: '%]' log: 'debug' show_errors: 'true' plack_middlewares: Debug: - panels - - Memory - DBITrace - Timer Here's the error message when I plackup $ plackup -S Starman --workers=10 -p 5000 -a bin/app.pl Error while loading bin/app.pl: Unable to parse the configuration file: /opt/foobar/webapps/cemconf/config.yml: Died at /usr/local/share/perl/5.10.0/YAML/Loader.pm line 676, <$IN> line 1. at /usr/local/share/perl/5.10.0/Dancer/Config.pm line 177 Dancer::Config::load_settings_from_yaml('/opt/foobar/webapps/cemconf/config.yml') called at /usr/local/share/perl/5.10.0/Dancer/Config.pm line 159 Dancer::Config::load('Dancer::Config') called at /usr/local/share/perl/5.10.0/Dancer.pm line 225 Dancer::import('Dancer') called at bin/app.pl line 2 Plack::Sandbox::bin_2fapp_2epl::BEGIN() called at /usr/local/share/perl/5.10.0/Dancer.pm line 2 eval {...} called at /usr/local/share/perl/5.10.0/Dancer.pm line 2 require bin/app.pl called at (eval 8) line 3 eval 'package Plack::Sandbox::bin_2fapp_2epl; { my $app = do $_file; if ( !$app && ( my $error = $@ || $! )) { die $error; } $app; } ;' called at /usr/local/share/perl/5.10.0/Plack/Util.pm line 108 Plack::Util::_load_sandbox('bin/app.pl') called at /usr/local/share/perl/5.10.0/Plack/Util.pm line 124 Plack::Util::load_psgi('bin/app.pl') called at /usr/local/share/perl/5.10.0/Plack/Runner.pm line 173 Plack::Runner::__ANON__() called at /usr/local/share/perl/5.10.0/Plack/Runner.pm line 23 Plack::Runner::__ANON__() called at /usr/local/share/perl/5.10.0/Plack/Runner.pm line 23 Plack::Runner::__ANON__() called at /usr/local/share/perl/5.10.0/Plack/Runner.pm line 23 Plack::Runner::__ANON__() called at /usr/local/share/perl/5.10.0/Plack/Runner.pm line 23 Plack::Runner::__ANON__() called at /usr/local/share/perl/5.10.0/Plack/Loader.pm line 56 Plack::Loader::preload_app('Plack::Loader=HASH(0x91d2550)', 'CODE(0x93054e0)') called at /usr/local/share/perl/5.10.0/Plack/Runner.pm line 253 Plack::Runner::run('Plack::Runner=HASH(0x9119bf0)') called at /usr/local/bin/plackup line 10 BEGIN failed--compilation aborted at bin/app.pl line 2. Thanks for any help! ------------------------------ todd.bruner(a)gmail.com

4 4

one route, different actions based on context
by Puneet Kishor 04 Jan '11

04 Jan '11

I am fairly certain I saw mention of this in the docs, but I can't find it now, so here goes -- The route `get '/all'` is invoked from the browser by going to http://application/foo/all. This causes a template to be rendered with all the "chrome" of the application, the buttons, the forms, the logos, bylines, etc. The crap that Edward Tufte refers to as "administrivia." This template is appropriately filled with data from a query for "all" foo widgets from the data store. Well, I also want `get '/all'` to callable from the command line. So, if the user types $ curl http://application/foo/all a json stream of all the foo widgets is returned. How do I accomplish the above? That is, have a single route called `get '/all'` and yet be able to call it either from a browser or from a command line but with different "stuff" returned? Seems like the ajax plugin is not the solution? or is it? I can't make out from its documentation. -- Puneet Kishor

4 8

sending file fails
by Puneet Kishor 04 Jan '11

04 Jan '11

This is driving me nuts, so I would appreciate any advice. Perhaps it is just a silly oversight on my part. punkish@Lucknow ~/Sites/carbonapp/public/images/models/2bau/years/8$pwd /Users/punkish/Sites/carbonapp/public/images/models/2bau/years/8 punkish@Lucknow ~/Sites/carbonapp/public/images/models/2bau/years/8$ls 2* 2010_nep.gif punkish@Lucknow ~/Sites/carbonapp/public/images/models/2bau/years/8$ So, as you can see, the following file exists /Users/punkish/Sites/carbonapp/public/images/models/2bau/years/8/2010_nep.gif My route is get '/:model_id/year/:year/zoom/:zoom/:output' => sub { my $model_id = params->{model_id}; my $year = params->{year}; my $zoom = params->{zoom}; my $output = params->{output}; # get model_file based on model_id my $dir = config->{'app_dir'} . '/' . config->{'img_dir'} . "/$model_file/years/$zoom"; my $img = $year . '_' . $output . '.gif'; debug("sending '$dir/$img'"); send_file("$dir/$img"); }; My development.log shows [37838] debug @0.008607> [hit #1] sending '/Users/punkish/Sites/carbonapp/public/images/models/2bau/years/8/2010_nep.gif' in /Users/punkish/Sites/carbonapp/lib/carbonapp/models.pm l. 106 and yet, I get the following error in my browser {"error":"No such file: `/Users/punkish/Sites/carbonapp/public/images/models/2bau/years/8/2010_nep.gif'"} Please point me to my stupid ways. -- Puneet Kishor Sent with Sparrow

2 2

Re: [Dancer-users] Dancer-users Digest, Vol 11, Issue 4
by Al Newkirk & Associates 03 Jan '11

03 Jan '11

okay cool, I think this reply should be cool...?!

2 1

Re: [Dancer-users] Dancer-users Digest, Vol 11, Issue 2
by awnstudio＠gmail.com 03 Jan '11

03 Jan '11

Your Rewrite is using .cgi instead of .fcgi. Also it doesn't have a DirectoryIndex. Did you know you could do ... <Files "app"> SetHandler cgi-script </Files> Which in effect treats "app" as a file not a folder :) No more need for symlink ln -s ... - Your truely Sent from my Verizon Wireless BlackBerry -----Original Message----- From: dancer-users-request(a)perldancer.org Sender: dancer-users-bounces(a)perldancer.org Date: Sun, 02 Jan 2011 12:00:02 To: <dancer-users(a)perldancer.org> Reply-To: dancer-users(a)perldancer.org Subject: Dancer-users Digest, Vol 11, Issue 2 Send Dancer-users mailing list submissions to dancer-users(a)perldancer.org To subscribe or unsubscribe via the World Wide Web, visit http://www.backup-manager.org/cgi-bin/listinfo/dancer-users or, via email, send a message with subject or body 'help' to dancer-users-request(a)perldancer.org You can reach the person managing the list at dancer-users-owner(a)perldancer.org When replying, please edit your Subject line so it is more specific than "Re: Contents of Dancer-users digest..." Today's Topics: 1. Re: CGI/FCGI deployment & uri_for (Naveed Massjouni) ---------------------------------------------------------------------- Message: 1 Date: Sat, 1 Jan 2011 06:38:50 -0500 From: Naveed Massjouni <naveedm9(a)gmail.com> Subject: Re: [Dancer-users] CGI/FCGI deployment & uri_for To: Robert Olson <bob(a)rdolson.org> Cc: Dancer-users(a)perldancer.org Message-ID: <AANLkTin+b=YygZ=Qw0HbH1cqVRNo+i6+jdzx1k0P00e-(a)mail.gmail.com> Content-Type: text/plain; charset=ISO-8859-1 On Tue, Dec 28, 2010 at 5:53 PM, Robert Olson <bob(a)rdolson.org> wrote: > It seems I must be doing something dumb, but I'm not seeing it. > > I'm putting up a dancer app for a domain hosted at dreamhost. I created the > app using the latest 1.2 dancer, and changed the webroot for the domain to > appdir/public using the DH config panel. The following .htaccess is in > place: > > AddHandler fastcgi-script .fcgi > Options +FollowSymLinks +ExecCGI > > RewriteEngine On > RewriteCond %{REQUEST_FILENAME} !-f > RewriteRule (.*) /dispatch.cgi/$1 [QSA,L] > > Access to the routes works like a champ (including > Net::Google::FederatedLogin authentication which I may turn into a plugin; > also thinking about a CGI::FormBuilder plugin like Catalyst's). > > However, uri_for returns a path that includes the rewritten path with > dispatch.cgi in it: > > uri_for("/dog") => http://<domain>/dispatch.cgi/dog > > I'm assuming it should instead return http://<domain>/dog. > > This seems like something there is an easy solution to, but I'm not finding > it. Does anyone have any advice? A simple thing that I do when deploying a CGI app is using a soft link instead of a rewrite rule: cd /var/www/cgi-bin # your system's cgi-bin may be at a different location ln -s /path/to/myapp/public/dispatch.cgi myapp You need to tell your web server to treat '^/myapp' as a cgi script. Then you can visit http://<domain>/myapp and it will be the root of your app. uri_for("/dog") should then return http://<domain>/myapp/dog I usually deploy as CGI for playing around or for development. When deploying for production, I use FastCGI, and then it is easy to mount your app to '/' or '/whatever'. -Naveed > > Thanks, > --bob > _______________________________________________ > Dancer-users mailing list > Dancer-users(a)perldancer.org > http://www.backup-manager.org/cgi-bin/listinfo/dancer-users > ------------------------------ _______________________________________________ Dancer-users mailing list Dancer-users(a)perldancer.org http://www.backup-manager.org/cgi-bin/listinfo/dancer-users End of Dancer-users Digest, Vol 11, Issue 2 *******************************************

2 1