Hey folks, I''m working on the Mongrel book with Zed, and wanted to get some feedback from the core users (this list) about how they use Mongrel. That sounds a bit vague, but I''m interested in hearing things about frustrating problems / workaround, preferred configurations, if you have a particular way you set up / store your config files when developing / deploying an app, for those of you running multiple handlers, if you have had to deal with any subtle issues, if there were any gotchas you encountered along the way from the docs that could be fixed (in the docs) or improved, any tricks you figured out along the way, etc. Any info will be credited properly. Thanks in advance. Matt ==================Matt Pelletier EastMedia t. 212-967-4239 f. 212-967-4257 m. 917-902-2525 w. www.eastmedia.com
Well, right now I use mongrel for deploying my live rails apps. Pretty standard I guess. I''ve currently been using mongrel_cluster, but I would like to eventually use some kind of process monitoring (monit?). I have yet to really look into it though. The book really should cover how to use mongrel with capistrano since it seems to be the "standard" deployment method. I''m also working on an application that would require to track bandwidth sent out per user (members upload files and get x amount of MB of bandwidth a month). I''m thinking of writing a mongrel handler and doing it like that, but I''m not sure if that''s the best way or not. -carl On 9/20/06, Matt Pelletier <matt at eastmedia.com> wrote:> Hey folks, > > I''m working on the Mongrel book with Zed, and wanted to get some > feedback from the core users (this list) about how they use Mongrel. > That sounds a bit vague, but I''m interested in hearing things about > frustrating problems / workaround, preferred configurations, if you > have a particular way you set up / store your config files when > developing / deploying an app, for those of you running multiple > handlers, if you have had to deal with any subtle issues, if there > were any gotchas you encountered along the way from the docs that > could be fixed (in the docs) or improved, any tricks you figured out > along the way, etc. Any info will be credited properly. > > Thanks in advance. > Matt > > ==================> Matt Pelletier > EastMedia > t. 212-967-4239 > f. 212-967-4257 > m. 917-902-2525 > w. www.eastmedia.com > > > _______________________________________________ > Mongrel-users mailing list > Mongrel-users at rubyforge.org > http://rubyforge.org/mailman/listinfo/mongrel-users >
> I''m working on the Mongrel book with Zed, and wanted to get some > feedback from the core users (this list) about how they use Mongrel. > That sounds a bit vague, but I''m interested in hearing things about > frustrating problems / workaround, preferred configurations, if you > have a particular way you set up / store your config files when > developing / deploying an app, for those of you running multiple > handlers, if you have had to deal with any subtle issues, if there > were any gotchas you encountered along the way from the docs that > could be fixed (in the docs) or improved, any tricks you figured out > along the way, etc. Any info will be credited properly.The only issue we''ve had is related to memcache connections not closing when you do a cluster:restart. See: http://rubyforge.org/tracker/index.php?func=detail&aid=5268&group_id=1306&atid=5145 Zed said this is a rails problem and I believe him. I imagine if I spent some time looking at it some more I''d figure out how to close the memcache connections within rails, but at the time we simply made the decision to hard restart mongrel (and apache) when necessary. Reason I mention this is the docs I was able to find about setting up memcache for use as fragment/session caching don''t mention this as an issue, and it bit us... Actually I guess there is one other issue we''ve noticed... on a heavily hit server, restarting just mongrel, apache seems to take a little while to "find" it again. So we restart apache after restarting mongrel. Super minor issues all things considered :-) We were restarting lighttpd/fastcgi via cron, hourly... -philip
Hey, We use the following elements for our deployment: 1. A hardware load balancer with two servers as backends, each running: 2. One lighttpd instance, routed to a pound acting as a load balancer, each with: 3. Two mongrel instances. 4. A monit process on each server monitoring pound, lighty & mongrel. 5. A separate database server with monit running on that too. In addition, adding Capistrano''s cap stop and start_mongrel_cluster magic scripts to the way we deploy. Mongrel''s been amazing. Vish On 9/21/06, Carl Lerche <carl.lerche at gmail.com> wrote:> > Well, right now I use mongrel for deploying my live rails apps. Pretty > standard I guess. I''ve currently been using mongrel_cluster, but I > would like to eventually use some kind of process monitoring (monit?). > I have yet to really look into it though. The book really should cover > how to use mongrel with capistrano since it seems to be the "standard" > deployment method. I''m also working on an application that would > require to track bandwidth sent out per user (members upload files and > get x amount of MB of bandwidth a month). I''m thinking of writing a > mongrel handler and doing it like that, but I''m not sure if that''s the > best way or not. > > -carl > > On 9/20/06, Matt Pelletier <matt at eastmedia.com> wrote: > > Hey folks, > > > > I''m working on the Mongrel book with Zed, and wanted to get some > > feedback from the core users (this list) about how they use Mongrel. > > That sounds a bit vague, but I''m interested in hearing things about > > frustrating problems / workaround, preferred configurations, if you > > have a particular way you set up / store your config files when > > developing / deploying an app, for those of you running multiple > > handlers, if you have had to deal with any subtle issues, if there > > were any gotchas you encountered along the way from the docs that > > could be fixed (in the docs) or improved, any tricks you figured out > > along the way, etc. Any info will be credited properly. > > > > Thanks in advance. > > Matt > > > > ==================> > Matt Pelletier > > EastMedia > > t. 212-967-4239 > > f. 212-967-4257 > > m. 917-902-2525 > > w. www.eastmedia.com > > > > > > _______________________________________________ > > Mongrel-users mailing list > > Mongrel-users at rubyforge.org > > http://rubyforge.org/mailman/listinfo/mongrel-users > > > _______________________________________________ > Mongrel-users mailing list > Mongrel-users at rubyforge.org > http://rubyforge.org/mailman/listinfo/mongrel-users >-------------- next part -------------- An HTML attachment was scrubbed... URL: http://rubyforge.org/pipermail/mongrel-users/attachments/20060921/26a06b8c/attachment-0001.html
On 9/20/06, Matt Pelletier <matt at eastmedia.com> wrote:> I''m working on the Mongrel book with Zed, and wanted to get some > feedback from the core users (this list) about how they use Mongrel. > That sounds a bit vague, but I''m interested in hearing things about > frustrating problems / workaround, preferred configurations, if you > have a particular way you set up / store your config files when > developing / deploying an app, for those of you running multiple > handlers, if you have had to deal with any subtle issues, if there > were any gotchas you encountered along the way from the docs that > could be fixed (in the docs) or improved, any tricks you figured out > along the way, etc. Any info will be credited properly.I have just recently (in the last two weeks) put working Mongrel support into IOWA, and I as well as some users are already using it as a deployment platform for sites and apps even though I''ve not done a new production release with the support, yet. I am also using the trie classifier library in one of the dispatchers that I am bundling with IOWA (dispatchers in IOWA are modular, so one can pick from a few in the distribtution, or just plug in something custom to do what one needs). Performance is comparable to using apache2 with the mod_ruby handler for communication, and is close to fcgi performance, but configuration and testing is much simpler, so it is a definite win for me. Kirk Haines
Vishnu, I''m curious to why you''re using lighttpd and pound instead of just pound? I don''t know much about either so the answer might be obvious. ed On 9/21/06, Vishnu Gopal <g.vishnu at gmail.com> wrote:> > Hey, > > We use the following elements for our deployment: > > 1. A hardware load balancer with two servers as backends, each running: > 2. One lighttpd instance, routed to a pound acting as a load balancer, > each with: > 3. Two mongrel instances. > 4. A monit process on each server monitoring pound, lighty & mongrel. > 5. A separate database server with monit running on that too. > > In addition, adding Capistrano''s cap stop and start_mongrel_cluster magic > scripts to the way we deploy. > > Mongrel''s been amazing. > Vish > > On 9/21/06, Carl Lerche <carl.lerche at gmail.com> wrote: > > > > Well, right now I use mongrel for deploying my live rails apps. Pretty > > standard I guess. I''ve currently been using mongrel_cluster, but I > > would like to eventually use some kind of process monitoring (monit?). > > I have yet to really look into it though. The book really should cover > > how to use mongrel with capistrano since it seems to be the "standard" > > deployment method. I''m also working on an application that would > > require to track bandwidth sent out per user (members upload files and > > get x amount of MB of bandwidth a month). I''m thinking of writing a > > mongrel handler and doing it like that, but I''m not sure if that''s the > > best way or not. > > > > -carl > > > > On 9/20/06, Matt Pelletier < matt at eastmedia.com> wrote: > > > Hey folks, > > > > > > I''m working on the Mongrel book with Zed, and wanted to get some > > > feedback from the core users (this list) about how they use Mongrel. > > > That sounds a bit vague, but I''m interested in hearing things about > > > frustrating problems / workaround, preferred configurations, if you > > > have a particular way you set up / store your config files when > > > developing / deploying an app, for those of you running multiple > > > handlers, if you have had to deal with any subtle issues, if there > > > were any gotchas you encountered along the way from the docs that > > > could be fixed (in the docs) or improved, any tricks you figured out > > > along the way, etc. Any info will be credited properly. > > > > > > Thanks in advance. > > > Matt > > > > > > ==================> > > Matt Pelletier > > > EastMedia > > > t. 212-967-4239 > > > f. 212-967-4257 > > > m. 917-902-2525 > > > w. www.eastmedia.com > > > > > > > > > _______________________________________________ > > > Mongrel-users mailing list > > > Mongrel-users at rubyforge.org > > > http://rubyforge.org/mailman/listinfo/mongrel-users > > > > > _______________________________________________ > > Mongrel-users mailing list > > Mongrel-users at rubyforge.org > > http://rubyforge.org/mailman/listinfo/mongrel-users > > > > > _______________________________________________ > Mongrel-users mailing list > Mongrel-users at rubyforge.org > http://rubyforge.org/mailman/listinfo/mongrel-users > >-- Ed Hickey Developer Litmus Media 816-533-0409 ehickey at litmusmedia.com A Member of Think Partnership, Inc www.ThinkPartnership.com Amex ticker symbol: THK -------------- next part -------------- An HTML attachment was scrubbed... URL: http://rubyforge.org/pipermail/mongrel-users/attachments/20060921/3e494ab1/attachment-0001.html
On 9/21/06, Philip Hallstrom <mongrel at philip.pjkh.com> wrote:> > > I''m working on the Mongrel book with Zed, and wanted to get some > > feedback from the core users (this list) about how they use Mongrel. > > That sounds a bit vague, but I''m interested in hearing things about > > frustrating problems / workaround, preferred configurations...We''re not doing anything too unique, but: 3 web servers and w/ a DB cluster on the backend: -F5 load balancer in front of 2 of the web servers (no software balancer) -the third server is running Apache as a balancer -each server has10 mongrels running one small problem i ran into the first time i installed mogrel was that i didn''t have gcc installed on the machine. mongrel installed without error, but when i tried to start it, i got cryptic errors (well, at least they were cryptic to me). after some googling i found out it was because of the missing complier so maybe that could be checked upon gem installation? ...or at least in the Readme (which it may well be by now). we''re having some other intermittent speeds issues, but i believe it''s a DB/Rails problem and can''t blame mongrel. we''re in beta and consistently serving a couple requests per second all day long without real issue. not bad indeed. i have another setup with nginx balancing 10 mongrels, but that''s too new to really comment on it much. regarding docs - there wasn''t any when i first installed mongrel. about the only thing i could find (of much substance) was a howto on some blog...seemed to work well enough. Zed''s new docs look very helpful though. i''ll try to supply some more useful info as our applications progress and we run into more obstacles. ed -------------- next part -------------- An HTML attachment was scrubbed... URL: http://rubyforge.org/pipermail/mongrel-users/attachments/20060921/d28f0c63/attachment.html
Hey Ed, On 9/22/06, Ed Hickey <bassnode at gmail.com> wrote:> > [..]-each server has10 mongrels running > [..] >One quick question: What''s the configuration of the servers you are running on? Thanks! Vishnu -------------- next part -------------- An HTML attachment was scrubbed... URL: http://rubyforge.org/pipermail/mongrel-users/attachments/20060922/80b1410b/attachment.html
On 9/21/06, Ed Hickey <bassnode at gmail.com> wrote:> We''re not doing anything too unique, but: > > 3 web servers and w/ a DB cluster on the backend: > -F5 load balancer in front of 2 of the web servers (no software balancer) > -the third server is running Apache as a balancer > -each server has10 mongrels running > > we''re having some other intermittent speeds issues, but i believe it''s a > DB/Rails problem and can''t blame mongrel. we''re in beta and consistently > serving a couple requests per second all day long without real issue. not > bad indeed.Is your application _really_ computation intensive? It boggles my mind that something running 3 web servers (single processor boxes?) with a separate db cluster (2 servers?) and 30 processes to handle it all could be having intermittent speed issues while serving only a couple requests per second. Kirk Haines
The deal is that it''s 1 Rails app that encompasses all the facets of an web advertising system (similar to Google''s Adwords). There are 3 different ''faces''. One serves ads in XML, one is an admin box and one handles clicks, exclusively. The 3 boxes aren''t connected in anyway other than the database cluster. Do we actually _need_ 10 mongrels running on each box? I can''t really say yet. We are expecting it to be quite a traffic-heavy system so we''re preparing... That said, we''re still in beta and testing the limits. I can say that the 10 mongrels aren''t weighing the servers down at all. The only thing that gets a little (memory) hungry are the Mysql processes which is what I think could be part of our intermittant speed problem. Sometimes they spiral up to 40MB a piece...and spawn into 30 different processes. We are using MySQL clustering (which is new to us) and it behaves much differently than standard MyISAM/InnoDB tables so we''re still sorting that out. I''m curious as to what you mean when when you say "all could be having intermittent speed issues while serving only a couple requests per second.". Were you assuming all the servers were clustered together? If not, how could they be having speed issues? Any input would be helpful. ed On 9/22/06, Kirk Haines <wyhaines at gmail.com> wrote:> > On 9/21/06, Ed Hickey <bassnode at gmail.com> wrote: > > > We''re not doing anything too unique, but: > > > > 3 web servers and w/ a DB cluster on the backend: > > -F5 load balancer in front of 2 of the web servers (no software > balancer) > > -the third server is running Apache as a balancer > > -each server has10 mongrels running > > > > Is your application _really_ computation intensive? It boggles my > mind that something running 3 web servers (single processor boxes?) > with a separate db cluster (2 servers?) and 30 processes to handle it > all could be having intermittent speed issues while serving only a > couple requests per second.Kirk Haines>-- Ed Hickey Developer Litmus Media 816-533-0409 ehickey at litmusmedia.com A Member of Think Partnership, Inc www.ThinkPartnership.com Amex ticker symbol: THK -------------- next part -------------- An HTML attachment was scrubbed... URL: http://rubyforge.org/pipermail/mongrel-users/attachments/20060922/64b5e08a/attachment-0001.html
On 9/22/06, Vishnu Gopal <g.vishnu at gmail.com> wrote:> > Hey Ed, > > On 9/22/06, Ed Hickey <bassnode at gmail.com> wrote: > > > > [..]-each server has10 mongrels running > > [..] > > > > One quick question: What''s the configuration of the servers you are > running on? > > Thanks! > Vishnu >Vishnu, The servers are Dual 2.8Ghz Xeons w/ 2GB RAM CentOS 3.6 (64bit) MySQL 5.0 (NDBCluster) Rais 1.16 Mongrel 0.3.13.4 Apache 2.2 (only on web frontend box) Apache is setup with mod_proxy so we can serve static content on the admin box. The other 2 boxes only serve XML and/or forward requests, so they are fronted by the F5 load balancers. All 3 boxes operate independently. They talk via the DB. I''m not sure that answers your question...? -- ed -------------- next part -------------- An HTML attachment was scrubbed... URL: http://rubyforge.org/pipermail/mongrel-users/attachments/20060922/7bcdc5c0/attachment.html