hi all, I''m having a lot of trouble with internal server errors, zombie fastcgi processes, crashes and whole trees of zombie apache processes: for example: [Thu May 19 18:44:39 MDT 2005] Dispatcher failed to catch: exit (SystemExit) /usr/lib/ruby/1.8/fcgi.rb:10:in `exit'' /usr/lib/ruby/1.8/fcgi.rb:10 /usr/lib/ruby/1.8/fcgi.rb:10:in `call'' /usr/lib/ruby/1.8/fcgi.rb:597:in `each'' /usr/lib/ruby/1.8/fcgi.rb:597:in `each_cgi'' /var/www/domains/site.example.com/beta/site/trunk/public/dispatch.fcgi: 18 FCGI process 13825 killed by this error Is there a known-good production quality setup for rails under FastCGI? What about mod_ruby? I''ve seen DHH''s suggestion: <IfModule mod_fastcgi.c> FastCgiIpcDir /tmp/fcgi_ipc FastCgiServer /app/public/dispatch.fcgi -initial-env RAILS_ENV=production -processes 15 -idle-timeout 60 </IfModule> but this is not allowed in virtualhosts, which I must have use of. suggestions? I''m also getting this in my apache error logs: [Thu May 19 18:32:54 2005] [error] [client 66.92.16.4] FastCGI: incomplete headers (0 bytes) received from server "/var/www/domains/site.example.com/beta/site/trunk/public/ dispatch.fcgi" [Thu May 19 18:34:22 2005] [error] [client 66.92.16.4] FastCGI: comm with (dynamic) server "/var/www/domains/site.example.com/beta/site/trunk/public/ dispatch.fcgi" aborted: (first read) idle timeout (30 sec) [Thu May 19 18:34:22 2005] [error] [client 66.92.16.4] FastCGI: incomplete headers (0 bytes) received from server "/var/www/domains/site.example.com/beta/site/trunk/public/ dispatch.fcgi" [Thu May 19 18:38:17 2005] [error] [client 66.92.16.4] FastCGI: comm with (dynamic) server "/var/www/domains/site.example.com/beta/site/trunk/public/ dispatch.fcgi" aborted: (first read) idle timeout (30 sec) [Thu May 19 18:38:17 2005] [error] [client 66.92.16.4] FastCGI: incomplete headers (0 bytes) received from server "/var/www/domains/site.example.com/beta/site/trunk/public/ dispatch.fcgi" [Thu May 19 18:46:08 2005] [error] [client 66.92.16.4] FastCGI: server "/var/www/domains/site.example.com/beta/site/trunk/public/ dispatch.fcgi" stderr: /var/www/domains/site.example.com/beta/site/trunk/public/../config/..// app/helpers/user_helper.rb:7: warning: already initialized constant DEFAULT_HEAD_OPTIONS [Thu May 19 18:48:26 2005] [error] [client 66.92.16.4] FastCGI: server "/var/www/domains/site.example.com/beta/site/trunk/public/ dispatch.fcgi" stderr: /var/www/domains/site.example.com/beta/site/trunk/public/../config/..// app/helpers/user_helper.rb:7: warning: already initialized constant DEFAULT_HEAD_OPTIONS, referer: http://beta.site.example.com/company [Thu May 19 18:48:27 2005] [error] [client 66.92.16.4] FastCGI: server "/var/www/domains/site.example.com/beta/site/trunk/public/ dispatch.fcgi" stderr: /var/www/domains/site.example.com/beta/site/trunk/public/../config/..// app/helpers/user_helper.rb:7: warning: already initialized constant DEFAULT_HEAD_OPTIONS, referer: http://beta.site.example.com/company/management -- alex black, founder the turing studio, inc. 510.666.0074 root-16h2cdTTKgpzNNFeSAH1EA@public.gmane.org http://www.turingstudio.com 2600 10th street, suite 635 berkeley, ca 94710
On 5/19/05, alex black <enigma-16h2cdTTKgpzNNFeSAH1EA@public.gmane.org> wrote:> hi all, I''m having a lot of trouble with internal server errors, zombie > fastcgi processes, crashes and whole trees of zombie apache processes:I completely know what you''re talking about. Getting FastCGI to behave has been an uphill battle. I believe that I have almost won, but every once in a while I still get the dreaded ''Rails application failed to start'' error that nothing but a `apachectl restart` seems to fix. Sincerely, Tom Reinhart tom-V0YqjHVuocLQT0dZR+AlfA@public.gmane.org http://AllTom.com/
On Thu, 2005-05-19 at 21:31 -0400, Tom Reinhart wrote:> I completely know what you''re talking about. Getting FastCGI to behave > has been an uphill battle. I believe that I have almost won, but every > once in a while I still get the dreaded ''Rails application failed to > start'' error that nothing but a `apachectl restart` seems to fix.I went with lighttpd+FCGI and it is rock stable. If you really must use apache, how about making lighttpd listen on a different port (or installed on a different server altogether) and then use apache''s mod_proxy to sit between your user and lighttpd? Just an idea but should work.
can i suggest using apache as a front end and lighttpd as the rails server... i tried using fast cgi under apache with rails and had lots of problems... under lighttpd it has been good... only problem i had was that min proc and max proc either have to be equal or min-proc HAS to be one or I end up with zombie fast cgis alex black wrote:> hi all, I''m having a lot of trouble with internal server errors, > zombie fastcgi processes, crashes and whole trees of zombie apache > processes: > > for example: > > [Thu May 19 18:44:39 MDT 2005] Dispatcher failed to catch: exit > (SystemExit) > /usr/lib/ruby/1.8/fcgi.rb:10:in `exit'' > /usr/lib/ruby/1.8/fcgi.rb:10 > /usr/lib/ruby/1.8/fcgi.rb:10:in `call'' > /usr/lib/ruby/1.8/fcgi.rb:597:in `each'' > /usr/lib/ruby/1.8/fcgi.rb:597:in `each_cgi'' > > /var/www/domains/site.example.com/beta/site/trunk/public/dispatch.fcgi: > 18 > FCGI process 13825 killed by this error > > > Is there a known-good production quality setup for rails under > FastCGI? What about mod_ruby? I''ve seen DHH''s suggestion: > > <IfModule mod_fastcgi.c> > FastCgiIpcDir /tmp/fcgi_ipc > FastCgiServer /app/public/dispatch.fcgi -initial-env > RAILS_ENV=production -processes 15 -idle-timeout 60 > </IfModule> > > but this is not allowed in virtualhosts, which I must have use of. > > suggestions? > > > I''m also getting this in my apache error logs: > > [Thu May 19 18:32:54 2005] [error] [client 66.92.16.4] FastCGI: > incomplete headers (0 bytes) received from server > "/var/www/domains/site.example.com/beta/site/trunk/public/ dispatch.fcgi" > [Thu May 19 18:34:22 2005] [error] [client 66.92.16.4] FastCGI: comm > with (dynamic) server > "/var/www/domains/site.example.com/beta/site/trunk/public/ > dispatch.fcgi" aborted: (first read) idle timeout (30 sec) > [Thu May 19 18:34:22 2005] [error] [client 66.92.16.4] FastCGI: > incomplete headers (0 bytes) received from server > "/var/www/domains/site.example.com/beta/site/trunk/public/ dispatch.fcgi" > [Thu May 19 18:38:17 2005] [error] [client 66.92.16.4] FastCGI: comm > with (dynamic) server > "/var/www/domains/site.example.com/beta/site/trunk/public/ > dispatch.fcgi" aborted: (first read) idle timeout (30 sec) > [Thu May 19 18:38:17 2005] [error] [client 66.92.16.4] FastCGI: > incomplete headers (0 bytes) received from server > "/var/www/domains/site.example.com/beta/site/trunk/public/ dispatch.fcgi" > [Thu May 19 18:46:08 2005] [error] [client 66.92.16.4] FastCGI: > server "/var/www/domains/site.example.com/beta/site/trunk/public/ > dispatch.fcgi" stderr: > /var/www/domains/site.example.com/beta/site/trunk/public/../config/..// > app/helpers/user_helper.rb:7: warning: already initialized constant > DEFAULT_HEAD_OPTIONS > [Thu May 19 18:48:26 2005] [error] [client 66.92.16.4] FastCGI: > server "/var/www/domains/site.example.com/beta/site/trunk/public/ > dispatch.fcgi" stderr: > /var/www/domains/site.example.com/beta/site/trunk/public/../config/..// > app/helpers/user_helper.rb:7: warning: already initialized constant > DEFAULT_HEAD_OPTIONS, referer: http://beta.site.example.com/company > [Thu May 19 18:48:27 2005] [error] [client 66.92.16.4] FastCGI: > server "/var/www/domains/site.example.com/beta/site/trunk/public/ > dispatch.fcgi" stderr: > /var/www/domains/site.example.com/beta/site/trunk/public/../config/..// > app/helpers/user_helper.rb:7: warning: already initialized constant > DEFAULT_HEAD_OPTIONS, referer: > http://beta.site.example.com/company/management > > > > >_______________________________________________ Rails mailing list Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org http://lists.rubyonrails.org/mailman/listinfo/rails
I have one question on this set-up. which is what i mean using... the remote_ip ( of course ) is set to the proxying apache server... how do i get the actual ip... with a backend mod_perl server... i know how to do this... i jsut hook into apache early on and replace the proxy server ip with the real one... can someone tell me how to get the same results with lighttpd... one thing on the apache proxying to lighttpd... if you have control of the whole machine... for various security purpose... i would suggest just have lighttpd listen on 127.0.0.1 assuming its a single machine setup... then you can run mod_security on the front end apache and check all requests for common attacks whereas if you listen on something other than localhost, the lighttpd is exposed directly Meng Kuan wrote:>I went with lighttpd+FCGI and it is rock stable. If you really must use >apache, how about making lighttpd listen on a different port (or >installed on a different server altogether) and then use apache''s >mod_proxy to sit between your user and lighttpd? Just an idea but should >work. > > >_______________________________________________ >Rails mailing list >Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org >http://lists.rubyonrails.org/mailman/listinfo/rails > > >_______________________________________________ Rails mailing list Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org http://lists.rubyonrails.org/mailman/listinfo/rails
Meng Kuan wrote:> I went with lighttpd+FCGI and it is rock stable. If you really must use > apache, how about making lighttpd listen on a different port (or > installed on a different server altogether) and then use apache''s > mod_proxy to sit between your user and lighttpd? Just an idea but should > work.I am not the original poster; however, in my case I am required to use Windows Server 2003 as the deployment platform. I have wasted many hours trying to get lighttpd+FCGI running in a reliable manner under windows - and it has been worse than with Apache. This is something that is concerning me greatly as I had to lobby aggressively to get approval to do my project in rails. Not just my reputation - my actual employment is at stake. I have read DHH''s opinion of Windows. He is entitled to that opinion. All the same, for Rails to live up to the hype I believe it needs to deploy robustly not just on linux, BSD et al - but also windows. I am now lobbying for some freeBSD servers. I am not likely to get them as it is hard to overcome the inertia of a 3000 employee company spread across several continents. I can fake it for a while with webrick. In a few months there will approx 10,000 customers hitting the app. When that happens, I will *have* to deploy something better - even if it means writing a dedicated server for Ruby. (which I then won''t be able to release as it will have been written on company time) xev going to
On Thu, 2005-05-19 at 21:33 -0500, Xevious wrote:> I am not the original poster; however, in my case I am required to > use Windows Server 2003 as the deployment platform. I have wasted > many hours trying to get lighttpd+FCGI running in a reliable manner > under windows - and it has been worse than with Apache.How about giving coLinux a try? I''ve got it working on my home PC using a debian image and it feels just like a normal debian server to me. Installation is quite simple. I remember someone on this list mentioned previously to have gotten rails + lighttpd + fcgi working on colinux.
On Thu, 19 May 2005, Xevious wrote:> Meng Kuan wrote: >> I went with lighttpd+FCGI and it is rock stable. If you really must use >> apache, how about making lighttpd listen on a different port (or >> installed on a different server altogether) and then use apache''s >> mod_proxy to sit between your user and lighttpd? Just an idea but should >> work. > > I am not the original poster; however, in my case I am required to use > Windows Server 2003 as the deployment platform. I have wasted many hours > trying to get lighttpd+FCGI running in a reliable manner under windows - and > it has been worse than with Apache. > > This is something that is concerning me greatly as I had to lobby > aggressively to get approval to do my project in rails. Not just my > reputation - my actual employment is at stake. > > I have read DHH''s opinion of Windows. He is entitled to that opinion. All > the same, for Rails to live up to the hype I believe it needs to deploy > robustly not just on linux, BSD et al - but also windows. > > I am now lobbying for some freeBSD servers. I am not likely to get them as > it is hard to overcome the inertia of a 3000 employee company spread across > several continents. > > I can fake it for a while with webrick. In a few months there will approx > 10,000 customers hitting the app. When that happens, I will *have* to deploy > something better - even if it means writing a dedicated server for Ruby. > (which I then won''t be able to release as it will have been written on > company time) > > xevi read these threads from time to time and wonder: how many man hours have been spent battling the apache fastcgi module and posting to various lists? how much energy is spent asking __free__ software to behave on a black-box operating system that no open source developers can possibly totally understand? i''m not trying to be critical or flame anyone - but the amount of energy spent on this list alone complaining about the apache fastcgi module on this list alone is a least 1000 fold the amount of energy it would have taken to make the module rock solid. something to seriously consider is why a very simple open source c extension of the apache httpd server that is so good in theory but not quite perfect in implementation doesn''t have developers beating down the door to contribute? if you think that you are too busy making a living to spend time on such a task consider how valuable being able to crank out web apps faster than the next guy can be - isn''t it worth polishing the tools used to do so? again, i''m not trying to critisize anyone in particular at all - just pleading with the rails community to put their heads together and lend a hand to the software that makes their jobs more enjoyable. a chain is only as strong as it''s weakest link and that link appears to be the fastcgi module of the ubiquitous httpd server. kind regards. -a -- ==============================================================================| email :: ara [dot] t [dot] howard [at] noaa [dot] gov | phone :: 303.497.6469 | My religion is very simple. My religion is kindness. | --Tenzin Gyatso ===============================================================================
Just to chime in with those forced to use Windows: I''m also forced to use Win32 for my Rails deployment, as my company uses SQL Server and the SQL Server Adapter currently requires ADO ... So getting a production quality Win32 platform is a must! Neville -----Original Message----- From: rails-bounces-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org [mailto:rails-bounces-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org] On Behalf Of Xevious Sent: Friday, 20 May 2005 12:33 PM To: rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org Subject: Re: [Rails] fastcgi stability == crap Meng Kuan wrote:> I went with lighttpd+FCGI and it is rock stable. If you really must > use apache, how about making lighttpd listen on a different port (or > installed on a different server altogether) and then use apache''s > mod_proxy to sit between your user and lighttpd? Just an idea but > should work.I am not the original poster; however, in my case I am required to use Windows Server 2003 as the deployment platform. I have wasted many hours trying to get lighttpd+FCGI running in a reliable manner under windows - and it has been worse than with Apache. This is something that is concerning me greatly as I had to lobby aggressively to get approval to do my project in rails. Not just my reputation - my actual employment is at stake. I have read DHH''s opinion of Windows. He is entitled to that opinion. All the same, for Rails to live up to the hype I believe it needs to deploy robustly not just on linux, BSD et al - but also windows. I am now lobbying for some freeBSD servers. I am not likely to get them as it is hard to overcome the inertia of a 3000 employee company spread across several continents. I can fake it for a while with webrick. In a few months there will approx 10,000 customers hitting the app. When that happens, I will *have* to deploy something better - even if it means writing a dedicated server for Ruby. (which I then won''t be able to release as it will have been written on company time) xev going to _______________________________________________ Rails mailing list Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org http://lists.rubyonrails.org/mailman/listinfo/rails
On 5/19/05, Neville Burnell <Neville.Burnell-uEDVyssJ3mUpAS55Wn97og@public.gmane.org> wrote:> So getting a production quality Win32 platform is a must!Heh, I didn''t realize there was such a thing. -- Urban Artography http://artography.ath.cx
Rob Park wrote:> Heh, I didn''t realize there was such a thing.I have always despised windows (well, not always, I thought WfWG was neat when it first arrived). I started out on HPUX and hit Linux somewhere around 0.95a. Later I became a BSD snob and was quite happy. (grew to love Solaris also) At some point in time I realized there was no point in being an OS zealot. I would still rather work on unix. Even so, I can''t deny the fact that my employer has many large Windows servers running in a mission critical environment. And, we live and die by queries in SQL Server over large datasets - all together about 500TB (yes, *terabytes*) of RAID. No, we don''t reboot the servers daily. My point is, yes, Windows servers were once a joke. And, in some ways they still are (I think IIS is a joke). However, Windows Server is not the dog it used to be. I think it would be great mistake to say "too bad for you windows losers, if you want to really use Rails get an OS that doesn''t suck". If I was an ASP.Net advocate that is *exactly* what I would want to hear. xev
Meng Kuan wrote:> How about giving coLinux a try? I''ve got it working on my home PC using > a debian image and it feels just like a normal debian server to me. > Installation is quite simple. I remember someone on this list mentioned > previously to have gotten rails + lighttpd + fcgi working on colinux.Actually, I looked at it and have considered it. The issue for me is that I have to sell it to the Windows heads that staff the data operations division. Apache is going to be a big enough battle. (that is why I would love a working Windows port of lighttpd - no one around here has heard of it and I can call it an "accelerator". Once the app deploys I won''t even have admin rights on the production servers. xev
* Ara.T.Howard <Ara.T.Howard-32lpuo7BZBA@public.gmane.org> [0559 03:59]:> energy spent on this list alone complaining about the apache fastcgi module > on > this list alone is a least 1000 fold the amount of energy it would have > taken > to make the module rock solid.Not if you factor in learning C and getting your head round apaches source tree - I suspect the second task would take a lot longer than the first. Part of the problem is that mod_fastcgi seems to be a lot less used than (say) mod_proxy - I''d say rails has been exercising it a lot more than it has been in the last year or two. It''s quicker for most people to stick lighttpd on and get on with writing Ruby (as an aside, the lighttpd codebase has the advantage of being a lot smaller than apaches, and Jan seems to be pretty helpful in accepting ''mosty right'' patches and fixing common bogons). -- ''Blackmail''s such an ugly word. I prefer extortion. The x makes it sound cool.'' -- Bender Rasputin :: Jack of All Trades - Master of Nuns
> hi all, I''m having a lot of trouble with internal server errors, > zombie fastcgi processes, crashes and whole trees of zombie apache > processes:* Are you running FastCGI 0.8.6[1]? * Are you using the latest FastCGI dispatcher[2]? I''ve found that FastCGI is usually my own fault, but that FCGI was really bad at telling me about it in the past. But with the new bridge and dispatcher, you get pretty good errors thrown in log/ fastcgi.crash.log. SystemExit should surely be a thing of the past. [1] http://raa.ruby-lang.org/project/fcgi/ [2] http://dev.rubyonrails.com/file/trunk/railties/dispatches/ dispatch.fcgi?rev=1284&format=raw -- David Heinemeier Hansson http://www.loudthinking.com -- Broadcasting Brain http://www.basecamphq.com -- Online project management http://www.backpackit.com -- Personal information manager http://www.rubyonrails.com -- Web-application framework
On Fri, 20 May 2005, Dick Davies wrote:>> energy spent on this list alone complaining about the apache fastcgi module >> on this list alone is a least 1000 fold the amount of energy it would have >> taken to make the module rock solid. > > Not if you factor in learning C and getting your head round apaches source > tree - I suspect the second task would take a lot longer than the first.granted - but i assume enough people already know C.> Part of the problem is that mod_fastcgi seems to be a lot less used than > (say) mod_proxy - I''d say rails has been exercising it a lot more than it > has been in the last year or two.exactly right i''d say - and precisely why the rails community is poised to take it into the limelight as perl did for cgi. it just needs a few people to lend a helping hand it seems. this has been on my to-do list for two years but i''m not doing any web developement at the moment so it''s tough to prioritize. i''m just hoping my comments might inspire someone else ;-)> It''s quicker for most people to stick lighttpd on and get on with writing > Ruby (as an aside, the lighttpd codebase has the advantage of being a lot > smaller than apaches, and Jan seems to be pretty helpful in accepting ''mosty > right'' patches and fixing common bogons).even if this were true, and i''ve no experience with lighthttpd, i think operating on apache is mandatory. in any case i''m glad to see a viaable alternative does exist. cheers. -a -- ==============================================================================| email :: ara [dot] t [dot] howard [at] noaa [dot] gov | phone :: 303.497.6469 | My religion is very simple. My religion is kindness. | --Tenzin Gyatso ===============================================================================
I''m in the same boat as the original poster. When I first heard of rails and did a couple tutorials, I realized instantly that it was my new preferred development platform. I told everyone about it at my job, got them into it, and now we''re always talking/sharing code, etc. Recently, management decided to take interest, and allowed us to purchase a server for rails depolyment. Our back end is SQL Server. We have .NET and java apps deployed on Windows Server 2003 boxes. I don''t know why we haven''t migrated to linux/sql server, but I''m sure it has something to do with .net (I''m in the java camp). With rails, I have been unsuccessful with creating a stable server environment for a couple weeks now. I have the same problems with everyone else using apache+fastcgi. At home, I have a desktop with linux+lighttpd+fcgi. Things are working quite well, but I find it very troubling that this is pretty much our only option for a production level deployment. With the SQL Server database connection being native to Win32 only, I can''t even test my apps at work under my box at home. The sucky part is that I really want to help, but have no real background in these types of problems. Jin On 5/20/05, Ara.T.Howard <Ara.T.Howard-32lpuo7BZBA@public.gmane.org> wrote:> On Fri, 20 May 2005, Dick Davies wrote: > > >> energy spent on this list alone complaining about the apache fastcgi module > >> on this list alone is a least 1000 fold the amount of energy it would have > >> taken to make the module rock solid. > > > > Not if you factor in learning C and getting your head round apaches source > > tree - I suspect the second task would take a lot longer than the first. > > granted - but i assume enough people already know C. > > > Part of the problem is that mod_fastcgi seems to be a lot less used than > > (say) mod_proxy - I''d say rails has been exercising it a lot more than it > > has been in the last year or two. > > exactly right i''d say - and precisely why the rails community is poised to > take it into the limelight as perl did for cgi. it just needs a few people to > lend a helping hand it seems. this has been on my to-do list for two years > but i''m not doing any web developement at the moment so it''s tough to > prioritize. i''m just hoping my comments might inspire someone else ;-) > > > It''s quicker for most people to stick lighttpd on and get on with writing > > Ruby (as an aside, the lighttpd codebase has the advantage of being a lot > > smaller than apaches, and Jan seems to be pretty helpful in accepting ''mosty > > right'' patches and fixing common bogons). > > even if this were true, and i''ve no experience with lighthttpd, i think > operating on apache is mandatory. in any case i''m glad to see a viaable > alternative does exist. > > cheers. > > -a > -- > ==============================================================================> | email :: ara [dot] t [dot] howard [at] noaa [dot] gov > | phone :: 303.497.6469 > | My religion is very simple. My religion is kindness. > | --Tenzin Gyatso > ==============================================================================> > _______________________________________________ > Rails mailing list > Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org > http://lists.rubyonrails.org/mailman/listinfo/rails >
On Fri, 20 May 2005, Jin Lee wrote:> I''m in the same boat as the original poster. When I first heard of rails and > did a couple tutorials, I realized instantly that it was my new preferred > development platform. I told everyone about it at my job, got them into it, > and now we''re always talking/sharing code, etc. Recently, management > decided to take interest, and allowed us to purchase a server for rails > depolyment. > > Our back end is SQL Server. We have .NET and java apps deployed on Windows > Server 2003 boxes. I don''t know why we haven''t migrated to linux/sql server, > but I''m sure it has something to do with .net (I''m in the java camp). With > rails, I have been unsuccessful with creating a stable server environment > for a couple weeks now. I have the same problems with everyone else using > apache+fastcgi. > > At home, I have a desktop with linux+lighttpd+fcgi. Things are working quite > well, but I find it very troubling that this is pretty much our only option > for a production level deployment. With the SQL Server database connection > being native to Win32 only, I can''t even test my apps at work under my box > at home. > > The sucky part is that I really want to help, but have no real background in > these types of problems.that''s great - i think desire is about all it takes. for instance, testing and doccumentation are nearly always lacking contributors in any open source project and this is certainly something you could be doing if there were an effort - which there isn''t now. of course what this really needs is someone versed in windows since that''s where fastcgi seems to be having problems. a more fundemental issue though, and one i don''t know if anyone has answered is - is fastcgi un-stable on windows because of the module itself or is the apache impl on windows the source of problems? obviously this requires a reproducable test failure and something with a windows devlopment platform at their disposal. can anyone out there identify consistent sources of errors at least and distill to a minimum program (preferably pure fcgi - not rails) that will demonstrate the problems? is anyone with a windows development environment and c experience willing to help with debugging such a case? cheers. -a -- ==============================================================================| email :: ara [dot] t [dot] howard [at] noaa [dot] gov | phone :: 303.497.6469 | My religion is very simple. My religion is kindness. | --Tenzin Gyatso ===============================================================================
Ara.T.Howard wrote:> that''s great - i think desire is about all it takes.Desire is good. Having available hours in the day to get it done is even better. :) (barely have time to mow my yard)> for instance, testing and doccumentation are nearly always> lacking contributors in any open source project Documentation I can do. (Once I have something to document.)> - is fastcgi un-stable on windows because of the module itself or is the > apache impl on windows the source of problems?I will be spending today trying to get rails/fastcgi working under IIS. If that is successful, I should be able to start testing fastcgi under IIS vs Apache. xev
> Desire is good. Having available hours in the day to get > it done is even better. :) (barely have time to mow my yard)Thanks Xevious...that made me laugh thinking of how I was in the same situation. My neighbors are probably not happy with the looks of my yard right now. -- Andrew Stone
> energy spent on this list alone complaining about the apache fastcgi > module on > this list alone is a least 1000 fold the amount of energy it would > have taken > to make the module rock solid.the amount of effort it takes me to whine about hoe crap fcgi is with apache2 is something like 10-50 orders of magnitude _less_ work than making stable software.> theory but not quite perfect in implementation doesn''t have developers > beating > down the door to contribute? if you think that you are too busy > making a > living to spend time on such a task consider how valuable being able > to crank > out web apps faster than the next guy can be - isn''t it worth > polishing the > tools used to do so?absolutely. I don''t know C ;)> it''s weakest link and that link appears to be the fastcgi module of the > ubiquitous httpd server.Totally - but I think it will happen when there is enough serious production use of FCGI that someone is willing to just do it. I can''t, no one at my company can - we don''t hack C. best, _alex
> Not if you factor in learning C and getting your head round apaches > source tree > - I suspect the second task would take a lot longer than the first.yep.> > Part of the problem is that mod_fastcgi seems to be a lot less used > than (say) > mod_proxy - I''d say rails has been exercising it a lot more than it > has been in > the last year or two._definitely_> It''s quicker for most people to stick lighttpd on and get on with > writing Ruby > (as an aside, the lighttpd codebase has the advantage of being a lot > smaller > than apaches, and Jan seems to be pretty helpful in accepting ''mosty > right'' > patches and fixing common bogons).The problem with lhttpd is experience. _every_ client I have with *nix boxes knows how to admin apache at a basic level. they have never heard of lighttpd - not to say it isn''t great, but we need a stable apache setup. is mod_ruby crappy or something? _a
> * Are you running FastCGI 0.8.6[1]?yes. libfcgi-ruby1.8 package on debian, 0.8.6-1> * Are you using the latest FastCGI dispatcher[2]?I just got it. Wow, it looks a _lot_ more mature.> I''ve found that FastCGI is usually my own fault, but that FCGI was > really bad at telling me about it in the past.Heh, yeah - I think that''s what''s up here too.> But with the new bridge and dispatcher, you get pretty good errors > thrown in log/fastcgi.crash.log. SystemExit should surely be a thing > of the past.Ok, excellent. The new dispatch.fcgi is much bigger and better. Are there any other files I should manually upgrade? thanks much for your help. best, _alex
> versed in windows since that''s where fastcgi seems to be having > problems. aTo clarify, I''m on debian sarge I am a bigot (anti-windows) because of some ugly production experiences, but I don''t talk about it - use what you like, though I do tend to find things that expect to work with apache _much_ better on *ix. For the record DHH (david tell me to call you something else if that pisses you off ;) - seems to be one of the only people here with serious, relatively large scale deployment experience with rails. Please, please please, sit down and write a couple pages on the webserver config for basecamp or something, so we can see how you guys are doing it. I use basecamp on a daily basis - it''s always fast and it''s always up, so there is a config that works... _really_well_ - I (and apparently other people) just don''t have it. mod_php is easy: put into apache, turn it on, it works. I don''t mind something harder but there needs to be some guidance. thanks! _alex
> Please, please please, sit down and write a couple pages on the > webserver config for basecamp or something, so we can see how you guys > are doing it.Here''s something I picked up from #rubyonrails. But for the life of me I can''t remember who was writing it: Enterprise Ruby On Rails http://defendem.com/read/book/1 -- rick http://techno-weenie.net
> For the record DHH (david tell me to call you something else if > that pisses you off ;) - seems to be one of the only people here > with serious, relatively large scale deployment experience with rails.The boys at 43 Things are doing some fairly heavy lifting too.> Please, please please, sit down and write a couple pages on the > webserver config for basecamp or something, so we can see how you > guys are doing it.I wrote a 30-paged chapter on Deployment for the Agile Web Development with Rails book. It includes a lot of these pointers. It should hopefully be available for purchase in beta form next week.> I use basecamp on a daily basis - it''s always fast and it''s always > up, so there is a config that works... _really_well_ - I (and > apparently other people) just don''t have it.It''s running on the SVN version of Rails. There''s no extra, hidden source that we''re applying ;~). FastCGI did boggle us for a while until we figured out how to make it report sane error messages back. Then we realized it was usually just our own fault ;) -- David Heinemeier Hansson http://www.loudthinking.com -- Broadcasting Brain http://www.basecamphq.com -- Online project management http://www.backpackit.com -- Personal information manager http://www.rubyonrails.com -- Web-application framework _______________________________________________ Rails mailing list Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org http://lists.rubyonrails.org/mailman/listinfo/rails
On Saturday, May 21, 2005, 10:18:29 AM, David wrote:>> Please, please please, sit down and write a couple pages on the >> webserver config for basecamp or something, so we can see how you >> guys are doing it.> I wrote a 30-paged chapter on Deployment for the Agile Web > Development with Rails book. It includes a lot of these pointers. It > should hopefully be available for purchase in beta form next week.Perhaps you should make the essential information free - it seems to be a stumbling block to a few people.>> I use basecamp on a daily basis - it''s always fast and it''s always >> up, so there is a config that works... _really_well_ - I (and >> apparently other people) just don''t have it.> It''s running on the SVN version of Rails. There''s no extra, hidden > source that we''re applying ;~). FastCGI did boggle us for a while > until we figured out how to make it report sane error messages back. > Then we realized it was usually just our own fault ;)So...? How do you make it report sane error messages? Gavin
On 5/21/05, Rick Olson <technoweenie-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:> > Please, please please, sit down and write a couple pages on the > > webserver config for basecamp or something, so we can see how you guys > > are doing it. > > Here''s something I picked up from #rubyonrails. But for the life of > me I can''t remember who was writing it: > > Enterprise Ruby On Rails > http://defendem.com/read/book/1This is from bougyman. He''s got some great tips here, and he''s dealing with heavy load too.> > -- > rick > http://techno-weenie.net > _______________________________________________ > Rails mailing list > Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org > http://lists.rubyonrails.org/mailman/listinfo/rails >-- Cheers Koz
Heya :)> -----Original Message----- > From: rails-bounces-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org [mailto:rails- > bounces-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org] On Behalf Of Xevious > Sent: Thursday, May 19, 2005 11:51 PM > To: rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org > Subject: Re: [Rails] fastcgi stability == crap > > However, Windows Server is not the dog it used to be.Indeed. Windows 2003 Server is far from a dog. It is stable, secure and runs like a top. The latest versions of Exchange and SQL server are also extremely robust and stable. Those people who oppose Windows are welcome to do so - but if they fail to at least realize how far it has come they will be very unpleasantly surprised.> think it would be great mistake to say "too bad for you > windows losers, if you want to really use Rails get an > OS that doesn''t suck".One of the problems with Rails is that to use it and get community support you have to buy into a lot of associated dogma or the answer often is "your wrong, do it our way or you''re an idiot". I know of several shops who would love to try Rails, but they aren''t going to do so until a reliable, stable and IIS hosted version is available. It won''t do much good to try and blame the problem on IIS when both ASP.NET and PHP can be run under very heavy loads under IIS without any real problem.> If I was an ASP.Net advocate that is *exactly* what I would > want to hear.Ignoring the Windows OS is not really a great idea. Soulhuntre ---------- http://www.girl2.com - my girls http://www.the-estate.com - my legacy http://wiki.thegreybook.com - my project http://weblog.soulhuntre.com - my thoughts
> > I wrote a 30-paged chapter on Deployment for the Agile Web > > Development with Rails book. It includes a lot of these pointers. It > > should hopefully be available for purchase in beta form next week. > > Perhaps you should make the essential information free - it seems to > be a stumbling block to a few people. >Perhaps i''d like a pony> >> I use basecamp on a daily basis - it''s always fast and it''s always > >> up, so there is a config that works... _really_well_ - I (and > >> apparently other people) just don''t have it. > > > It''s running on the SVN version of Rails. There''s no extra, hidden > > source that we''re applying ;~). FastCGI did boggle us for a while > > until we figured out how to make it report sane error messages back. > > Then we realized it was usually just our own fault ;) > > So...? How do you make it report sane error messages?Like he said there is no hidden code on their site. If you use the dispatch.fcgi which comes with the latest rails you will get a log/fcgi.crash.log file in which you can find all the information -- Tobi http://www.snowdevil.ca - Snowboards that don''t suck http://www.hieraki.org - Open source book authoring http://blog.leetsoft.com - Technical weblog
> > > >One of the problems with Rails is that to use it and get community support >you have to buy into a lot of associated dogma or the answer often is "your >wrong, do it our way or you''re an idiot". > >I know of several shops who would love to try Rails, but they aren''t going >to do so until a reliable, stable and IIS hosted version is available. It >won''t do much good to try and blame the problem on IIS when both ASP.NET and >PHP can be run under very heavy loads under IIS without any real problem. > >What is involved in bringing Rails on IIS up to speed with other solutions?
>>One of the problems with Rails is that to use it and get community >> support >>you have to buy into a lot of associated dogma or the answer often is >> "your >>wrong, do it our way or you''re an idiot".That hasn''t been my experience with the Rails community at all. I''ve never had anyone tell me I''m an idiot ;) or even be particularly rude - I have got a lot of help and communicated a lot - all to my satisfaction. To a large extent that was part of my test of this framework, before I make a major platform decision - so I''m not saying it hasn''t happened, but I am confident saying that it''s not characteristic of the rails community. People are here because they want to get work done, and would like to use cool, efficient tools. Rails is a baby, it''s new, so it''s going to have growing pains, such is life. If you want vendor support use something that isn''t 10 seconds old. (Of course it probably won''t be very cool)...>>to do so until a reliable, stable and IIS hosted version is available. It >>won''t do much good to try and blame the problem on IIS when both ASP.NET >> and >>PHP can be run under very heavy loads under IIS without any real problem.Coming from a heavy unix background, and having terrible experiences with Win/IIS in production, I do understand this position even though I don''t agree with it. Anyway - yes, it would be a good thing to have ROCK SOLID configs available for apache and IIS for rails, same with lighttpd. _a
>> Perhaps you should make the essential information free - it seems to >> be a stumbling block to a few people.> Perhaps i''d like a ponyCome on, that''s a pretty bullshit response. If you''re going to advertise a framework as opensource, it''s necessary to provide the supporting information to run it in production, or people like me (that have the power to seriously contribute) will leave, quickly. I plan to buy the book, many of them, but the original request is spot on: whoever has experimented with FCGI and has stable production configs really should provide them - at the moment that''s 3 entities: 37signals, you, and 43things. There is no point in forcing the community to experiement for itself when existing configs work.>> So...? How do you make it report sane error messages? > > Like he said there is no hidden code on their site. If you use the > dispatch.fcgi which comes with the latest rails you will get a > log/fcgi.crash.log file in which you can find all the informationI am now, I''ll report back about the effectiveness of those messages. _a
michael-ik5EoZIp/sowyUHIeJf27w@public.gmane.org
2005-May-22 17:12 UTC
Re: fastcgi stability == crap, DHH
>> > I wrote a 30-paged chapter on Deployment for the Agile Web >> > Development with Rails book. It includes a lot of these pointers. It >> > should hopefully be available for purchase in beta form next week. >> >> Perhaps you should make the essential information free - it seems to >> be a stumbling block to a few people. >> > Perhaps i''d like a ponyThat really isn''t a valid response - I doubt Rails is going to get anywhere if the answer to, "How do I make my app fast/stable enough for prod.?", is, "Please pay for this information." On this subject, has anyone tried using fcgid? Is it any better than fcgi? Michael
> That really isn''t a valid response - I doubt Rails is going to get > anywhere if the answer to, "How do I make my app fast/stable enough for > prod.?", is, "Please pay for this information."Yep.> On this subject, has anyone tried using fcgid? Is it any better than fcgi?I don''t believe there is a debian package for it, so I haven''t tried it. I''ll look tomorrow. I''m inclined to believe DHH - that fcgi is fine but needs tweaking. I''d love to actually get _details_ on a setup that is working in a cluster behind a load balancer, though, and it''s definitely frustrating to know there are people running stable configs but aren''t providing them. thanks, _alex
On Sun, 2005-05-22 at 11:34 -0700, alex black wrote:> > That really isn''t a valid response - I doubt Rails is going to get > > anywhere if the answer to, "How do I make my app fast/stable enough for > > prod.?", is, "Please pay for this information." > > Yep. > > > On this subject, has anyone tried using fcgid? Is it any better than fcgi? > > I don''t believe there is a debian package for it, so I haven''t tried it. > I''ll look tomorrow.from apt-get: libapache2-mod-fcgid - an alternative module compat with mod_fastcgi Looks like it''s available for Sid and Sarge: http://packages.qa.debian.org/liba/libapache2-mod-fcgid.html I haven''t tried using is myself, so I can''t speak to it''s efficiency
On 5/23/05, alex black <enigma-16h2cdTTKgpzNNFeSAH1EA@public.gmane.org> wrote:> > That really isn''t a valid response - I doubt Rails is going to get > > anywhere if the answer to, "How do I make my app fast/stable enough for > > prod.?", is, "Please pay for this information." > > Yep.Have you guys read the URL provided in the second email in this thread? http://defendem.com/read/book/1> > On this subject, has anyone tried using fcgid? Is it any better than fcgi? > > I don''t believe there is a debian package for it, so I haven''t tried it. > I''ll look tomorrow. > > I''m inclined to believe DHH - that fcgi is fine but needs tweaking. I''d > love to actually get _details_ on a setup that is working in a cluster > behind a load balancer, though, and it''s definitely frustrating to know > there are people running stable configs but aren''t providing them. > > thanks, > > _alex > _______________________________________________ > Rails mailing list > Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org > http://lists.rubyonrails.org/mailman/listinfo/rails >-- Cheers Koz
According to Michael Koziarski:> Have you guys read the URL provided in the second email in this thread? > > http://defendem.com/read/book/1I''m trying :-) 500 - Internal Server Error -- Ollivier ROBERT -=- FreeBSD: The Power to Serve! -=- roberto-0kjVc+YyuDZX+h8frlqCcVAUjnlXr6A1@public.gmane.org Darwin snuadh.freenix.org Kernel Version 7.9.0: Wed Mar 30 20:11:17 PST 2005
On May 21, 2005, at 11:29 AM, alex black wrote:> whoever has experimented with FCGI and has stable production configs > really should provide them - at the moment that''s 3 entities: > 37signals, > you, and 43things. There is no point in forcing the community to > experiement for itself when existing configs work.On May 22, 2005, at 11:34 AM, alex black wrote:> I''d love to actually get _details_ on a setup that is working in a > cluster > behind a load balancer, though, and it''s definitely frustrating to > know > there are people running stable configs but aren''t providing them.I think the lack of specific answers is partially because it''s rather general question, "a cluster behind a load balancer" is a good first step towards something more specific. Even a "cluster" means different things to different people. I hear cluster and think of a 110 node rack in front of a 10+TB SAN. Alex, perhaps you can start a functional, case-based approach to this? Come up with some cases (test/mock if need be) and specific scenarios because that''s what is lacking. What''s the initial performance of your application under what conditions? What kind of throughput do you need? How many people is it going to serve? What should it scale to and how quickly? What can you afford? Are you clustering more for low latency/speed/load or do have a particular need for never being down (they are a little different). Sure there are little nuances that vary from server-to-server and some OSes in a fastcgi-driven setup but if you''ve done it in in perl or php, then the fastcgi part of it isn''t too different. And that said, "production scaling" and optimization is really the kind of problem that problem that 99% of app developers won''t be lucky enough to have. - Jason
On 5/22/05, Ollivier Robert <roberto-0kjVc+YyuDZX+h8frlqCcVAUjnlXr6A1@public.gmane.org> wrote:> According to Michael Koziarski: > > Have you guys read the URL provided in the second email in this thread? > > > > http://defendem.com/read/book/1 > > I''m trying :-) > > 500 - Internal Server Error >Same here.
On 5/23/05, Joe Van Dyk <joevandyk-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:> On 5/22/05, Ollivier Robert <roberto-0kjVc+YyuDZX+h8frlqCcVAUjnlXr6A1@public.gmane.org> wrote: > > According to Michael Koziarski: > > > Have you guys read the URL provided in the second email in this thread? > > > > > > http://defendem.com/read/book/1 > > > > I''m trying :-) > > > > 500 - Internal Server Error > > > > Same here.works for me. bug bougyman on the irc channel if you''re still having problems _______________________________________________> Rails mailing list > Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org > http://lists.rubyonrails.org/mailman/listinfo/rails >-- Cheers Koz
Heya :)> -----Original Message----- > From: rails-bounces-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org [mailto:rails- > bounces-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org] On Behalf Of alex black > Sent: Saturday, May 21, 2005 2:30 PM > To: Tobias Luetke; rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org > Subject: Re: [Rails] fastcgi stability == crap, DHH > > > Perhaps i''d like a pony > > Come on, that''s a pretty bullshit response. If you''re going to advertise a > framework as opensource, it''s necessary to provide the supporting > information to run it in production, or people like me (that have the > power to seriously contribute) will leave, quickly.Yeah, it really is a pretty crazy response. While I applaud the tightrope effort in trying to push rails for profit the trade offs will be dangerous ones. You can''t really say "adopt this free framework... but if you really want to learn how to use it then buy my book... you even get a set of steak knives!" Soulhuntre ---------- http://www.girl2.com - my girls http://www.the-estate.com - my legacy http://wiki.thegreybook.com - my project http://weblog.soulhuntre.com - my thoughts
> Yeah, it really is a pretty crazy response. While I applaud the > tightrope > effort in trying to push rails for profit the trade offs will be > dangerous > ones. You can''t really say "adopt this free framework... but if you > really > want to learn how to use it then buy my book... you even get a set > of steak > knives!"Rails has a true wealth of documentation in form of the API, the many tutorials, the wiki, bloggers, and more. Yes, that''s a scattered experience because collecting and vetting it all is hard and takes time. Commercially backed projects to remedy that are now entering the scene. The Agile Web Development with Rails is one of those projects and yes its for pay. But I''m rather insulted by the accusation that Rails is being pushed just so we can sell steak knives by withholding information. All the information to run a stable FCGI setup is out there. I posted links to the latest dispatch.fcgi that Basecamp and Backpack is running earlier in this thread. It''s in the SVN repository, so its hardly secret sauce. Dealing with FCGI on Windows is just not something any of the core contributors are working on, but happily it sounds like other people in this thread is pushing it. So please, tone down the outcry. It _sounds_ (I''m not saying is) a lot like wanting more for nothing. I poor endless hours into Rails where the product is instantly available from SVN and frequent releases. Others are busy filling in with free documentation. And others still are spending months compiling that coherent story (Dave Thomas). Choose what fits you well. I know as well as any that it can be frustrating to get technology working. Venting that frustration can be done in many ways, though. I''m suggesting that there may be better ways than what''s been exhibited in this thread. -- David Heinemeier Hansson http://www.loudthinking.com -- Broadcasting Brain http://www.basecamphq.com -- Online project management http://www.backpackit.com -- Personal information manager http://www.rubyonrails.com -- Web-application framework _______________________________________________ Rails mailing list Rails-1W37MKcQCpIf0INCOvqR/iCwEArCW2h5@public.gmane.org http://lists.rubyonrails.org/mailman/listinfo/rails
On 5/23/05, Soulhuntre <soulhuntre-xtZVmrYH4z1ZroRs9YW3xA@public.gmane.org> wrote:> Yeah, it really is a pretty crazy response. While I applaud the tightrope > effort in trying to push rails for profit the trade offs will be dangerous > ones. You can''t really say "adopt this free framework... but if you really > want to learn how to use it then buy my book... you even get a set of steak > knives!"It''s a non-issue. I managed to get stable fastcgi setup using only information that I found with google. This page was most helpful for me: http://bougyman.com/miscfiles/RailsonDebian.html (yeah, it''s for lighttpd and not apache, but lighttpd is where it''s at) -- Urban Artography http://artography.ath.cx
> So please, tone down the outcry. It _sounds_ (I''m not saying is) a > lot like wanting more for nothing. I poor endless hours into Rails > where the product is instantly available from SVN and frequent > releases. Others are busy filling in with free documentation. And > others still are spending months compiling that coherent story (Dave > Thomas). Choose what fits you well.Well said, and I for one am more than happy to purchase both the beta ebook and paper version when they become available. Writing accurate, coherent documentation takes a lot of time and I don''t expect it for free. Furthermore, I''m more than happy to indirectly support Rails development through my book purchases. And I think you will find most people on this list will perfectly mirror my sentiments - keep up the good work David! Ben
I think the "outcry" was primarily a reactionary response to the _pony_ line. Responses like the _pony_ line don''t help anything - they just piss people off. If newcomers feel an intense elitist attitude when they try to get started, they leave and go elsewhere (back to Java/Spring/Hibernate/whatever). Books are good though, and most everyone is willing to buy a book. MT
> So please, tone down the outcry. It _sounds_ (I''m not saying is) a lot > like wanting more for nothing. I poor endless hours into Rails where > the product is instantly available from SVN and frequent releases. > Others are busy filling in with free documentation. And others still > are spending months compiling that coherent story (Dave Thomas). Choose > what fits you well.The points that were raised should not be considered an "outcry", but rather as constructive criticism, if anything. The fcgi story is obviously a sore point for what seems to be more than a handful of rails users; pushing for them to use lighthttpd is just adding another "new" technology that users are going to have to persuade their server admins to adopt (a few seem to be already struggling with rails adoption). As for wanting more for nothing, I believe that 99% of us are more than happy to buy the book, but optimised fcgi configs should still be available online. It''s great that you spend so much time on rails, but then again, it is your baby, I''d be a bit worried if the creator of such an innovative framework didn''t spend most of his day working on it...> I know as well as any that it can be frustrating to get technology > working. Venting that frustration can be done in many ways, though. > I''m suggesting that there may be better ways than what''s been exhibited > in this thread.I think we should all make an effort to wiki any significant (or even non-significant) information that we discover in our own rails environments - I put up apache/fcgi/gentoo instructions a couple of months ago; I''m still running that setup, and it has never given me too much trouble (fcgi went on it''s first rampage last week, destroying all memory and cycles before it :P). Michael