Does anybody here had any luck on tracking down memory leaks on Rails apps? i've got a small application that no matter what i try to do,the mongrels processes keep growing and growing up.I've heard Evan made this gem called bleak house to help people track down memory leaks,but I couldnt get anything helpful from thoses graphs its generate.Does anybody have any suggestions? Thankz in advance!
on 2008-10-07 04:40
on 2008-10-09 10:00
Sorry I don't have an answer for you at the moment, but I would make the same request. I have recently started to experience serious memory leaks in my two live apps (running on different servers). I am using lighttpd and fastcgi on Fedora 8 and 9. One is using Rails 2.0.2,and one Rails 1.2.6 Ruby version 1.8.6 (2008-06-20 patchlevel 230) I also seem to get the memory leak on my development system running mongrel. I am currently investigating and will post back here with anything I come up with that may help. I only noticed the problem when I was investigating a hardware failure, but had been getting comments from users about slow responses for a little while. I am monitoring the process memory size and watch it grow every few requests. If anyone can help that would be great. Tonypm
on 2008-10-09 10:17
On 9 Oct 2008, at 08:59, tonypm wrote: > ruby 1.8.6 p230 is screwed and known for leaking and segfaulting. Fred
on 2008-10-12 18:50
Thanks Fred, Fedora updated Ruby to: ruby 1.8.6 (2008-08-11 patchlevel 287) [i386-linux] problems now all gone away - much relieved!! Tonypm
on 2008-10-14 04:08
Does patchlevel 111 have the same problem as well? I've been running ruby 1.8.6 patchlevel 111 on mac,but the processes grow slightly every request.On my production server running debian,I've got ruby 1.8.5 from 2006-08-25,and because I've got a lot of traffic on my production server,memory goes away really fast!I'll update to ruby 1.8.6 latest release and post the results here. Thankz!
on 2008-10-14 12:52
On 14 Oct 2008, at 03:07, Alex Gregianin wrote: > > Does patchlevel 111 have the same problem as well? > > I've been running ruby 1.8.6 patchlevel 111 on mac,but the > processes grow slightly every request.On my production server running > debian,I've got ruby 1.8.5 from 2006-08-25,and because I've got a lot > of traffic on my production server,memory goes away really fast!I'll > update to ruby 1.8.6 latest release and post the results here. > shouldn't do (p230 is where the bad stuff happens). It is perfectly possible to have a memory leak that is your fault - see if you can isolate it to a segment of code that leaks Fred
on 2008-10-17 02:21
We're running mongrels on Ubuntu with Ruby 1.8.6-p114, and memory of each process is slowly increasing, so we have to restart them every couple of days or so. Anybody else having to do this? I tried to use bleak-house, but that refused to work without sigsegv'ing. So I've written a simple action that walks the ObjectSpace and counts the number of objects per class and renders them. Nothing obvious there that would account for 20MB of leakage. When we start mongrel, they use 60MB quickly, and then slowly increase memory every request. When they're at 90MB or 100MB, we restart them (because we have 18 running on one box, and they use up lots of memory). Have some question to you guys: - How often do you restart mongrels? - How much memory does your mongrel orocess use (after a couple hours of uptime)? - What Ruby version is the most recommended? thx Andrej