I needed to restart my mongrel cluster today (updated a Rails controller). And my mongrels didn''t really come back up. They bind the port but don''t respond to any requests. I can''t find any log file that has useful information. These were running fine until I tried the restart. And I have not changed any software versions in a month. (And yes, I tried rolling back the controller change.) Where can I find more information about why this isn''t responding anymore? Ruby: ruby 1.8.5 (2006-08-25) [i386-linux] Mongrel: mongrel (1.0.1) Rails: 1.2.3 The only thing in the mongrel log is: ** Starting Mongrel listening at 192.168.19.116:3000 ** Changing group to staff. ** Changing user to mongrel. ** Loading additional MIME types from ** /srv/config/mongrel/extra_mime_types.yml ** Starting Rails with production environment... Wget connects and then hangs with: HTTP request sent, awaiting response... Top shows: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 2305 mongrel 25 0 38404 19m 4192 R 50.7 7.7 6:20.55 mongrel_rails 2308 mongrel 25 0 38396 19m 4188 R 49.3 7.7 6:26.68 mongrel_rails -- Cynthia Kiser
I hate mailing list threads that post a problem but not the solution, so I thought I would write back to say that the issue appears to have been some strangness with the Oracle libraries. Looking at our snapshotted volume, the libraries had not changed since installing about 2 months ago. However, for reasons that are still unclear, any attempt to connect to Oracle from that box looked at the sqlnet.ora file and then went into a strange polling state that consumed all available CPU. After getting some hint of this from stracing the hung mongrel process, we were able to reproduce the same problem using just sqlplus. So this was NOT a mongrel nor a rails issue. Perhaps even more mystriously, rebooting the server (a Xen VM running RHEL 5) fixed the problem. Thanks to the folks here and on the mongrel IRC room for all their suggestions on how to track this down. -- Cynthia Kiser
Cynthia, Thank you VERY much for the followup. ~Wayne. -------------- next part -------------- An HTML attachment was scrubbed... URL: http://rubyforge.org/pipermail/mongrel-users/attachments/20071130/b2db2429/attachment.html
> I needed to restart my mongrel cluster today (updated a Rails > controller). And my mongrels didn''t really come back up. They bind the > port but don''t respond to any requests. I can''t find any log file that > has useful information. These were running fine until I tried the > restart. And I have not changed any software versions in a month. (And > yes, I tried rolling back the controller change.)I had this problem on os x recently--if mongrel has been running ''awhile'' (i.e. answered a few requests), then i hit ctrl-c on it, it uses 100% cpu and never terminates. Very odd. Might be a ruby bug. Any thoughts? (this with mysql, leopard, mongrel 1.1.3). thanks. -Roger -- Posted via http://www.ruby-forum.com/.
On Feb 4, 2008 2:38 PM, Roger Pack <lists at ruby-forum.com> wrote:> > I needed to restart my mongrel cluster today (updated a Rails > > controller). And my mongrels didn''t really come back up. They bind the > > port but don''t respond to any requests. I can''t find any log file that > > has useful information. These were running fine until I tried the > > restart. And I have not changed any software versions in a month. (And > > yes, I tried rolling back the controller change.) > > > I had this problem on os x recently--if mongrel has been running > ''awhile'' (i.e. answered a few requests), then i hit ctrl-c on it, it > uses 100% cpu and never terminates. Very odd. Might be a ruby bug. > Any thoughts? > (this with mysql, leopard, mongrel 1.1.3). > thanks. > -Roger >I also have experienced this with Rails but not Merb so I''m guessing that is where the issue lie. ~Wayne -------------- next part -------------- An HTML attachment was scrubbed... URL: http://rubyforge.org/pipermail/mongrel-users/attachments/20080205/5843b919/attachment.html