Zachary Powell
2007-May-07 00:15 UTC
[Mongrel] Mongrel and EBADF error consuming resources
In the last few weeks I''ve noticed the load average spiking up for 5 minutes at a time (2-3.5 on a system that usually runs around 0.5-1.5). Looking closer I found that one mongrel process and litespeed are at the top, and using strace I''ve found that they''re passing the file below, seems to get stuck on single file/EBADF error. It happens with a number of different urls, and they''re all caches_page so I don''t think it has anything to do with my rails code. Also, the mongrel process is still usable during this time (I can request pages from it and the loop below continues uneffected, so its not one request locking up the process). Any ideas on how I can straighten this out? thanks, Zach (mongrel (1.0.1, cluster 0.2.0, rails 1.1.16, ruby 1.8.4 on fedora) read(188, "GET /flower_delivery/florists_in"..., 16384) = 471 close(188) = 0 close(188) = -1 EBADF (Bad file descriptor) read(188, "GET /flower_delivery/florists_in"..., 16384) = 471 close(188) = 0 close(188) = -1 EBADF (Bad file descriptor) read(188, "GET /flower_delivery/florists_in"..., 16384) = 471 close(188) = 0 close(188) = -1 EBADF (Bad file descriptor) read(188, "GET /flower_delivery/florists_in"..., 16384) = 471 close(188) = 0 close(188) = -1 EBADF (Bad file descriptor) --- SIGVTALRM (Virtual timer expired) @ 0 (0) --- read(188, "GET /flower_delivery/florists_in"..., 16384) = 471 close(188) = 0 close(188) = -1 EBADF (Bad file descriptor) read(188, "GET /flower_delivery/florists_in"..., 16384) = 471 close(188) = 0 close(188) = -1 EBADF (Bad file descriptor) read(188, "GET /flower_delivery/florists_in"..., 16384) = 471 close(188) = 0 close(188) = -1 EBADF (Bad file descriptor) read(188, "GET /flower_delivery/florists_in"..., 16384) = 471 close(188) = 0 close(188) = -1 EBADF (Bad file descriptor) read(188, "", 16384) = 0 close(188) = 0 close(188) = -1 EBADF (Bad file descriptor)
On May 6, 2007, at 5:15 PM, Zachary Powell wrote:> In the last few weeks I''ve noticed the load average spiking up for 5 > minutes at a time (2-3.5 on a system that usually runs around > 0.5-1.5). Looking closer I found that one mongrel process and > litespeed are at the top, and using strace I''ve found that they''re > passing the file below, seems to get stuck on single file/EBADF error.I''ve seen this too but I haven''t found the cause yet. -- miles "If money played an instrument it would be the saxophone." - Achewood