I’ve got a problem with timing out a C RFC call.
Essentially, I’ve got code that initializes and executes an RFC to a
remote server that, depending on the server, could take a significant
amount of time to return a response.
To minimize end-user impact, I wrapped the call with a timeout of 20
seconds.
When I executed a test, it didn’t appear to do anything as the process
still took 120-130 seconds to complete.
After digging around a bit, I discovered this writeup about timeout and
system calls
(http://adamblog.heroku.com/past/2008/6/17/battling_wedged_mongrels_with_a/).
I immediately installed SystemTimer
(http://ph7spot.com/articles/system_timer) and tried again:
begin
require ‘system_timer’
MyTimer = SystemTimer
rescue LoadError
require ‘timeout’
MyTimer = Timeout
end
def invoke(rfc)
MyTimer.timeout_after(20.seconds) do
rfc.invoke
end
rescue Timeout::Error
raise Timeout::Error, “RFC Call timed out after 20 seconds”
end
This time, it executed and waited the full length of time (120-130
seconds). 20 seconds AFTER it completed, I then saw the “RFC Call timed
out after 20 seconds” error written to the logs.
After some more investigation, I read about Terminator and the use of
new threads to kill stuck/long-running threads.
I installed it and updated my above block of code to use the Terminator
syntax.
This time when I ran it, a view of the process list showed that the new
thread got created with a timer of 20 seconds. 20 seconds into
processing, the thread showed as and nothing happened to the
existing thread… that is until it completed. Once the process
finished running, the mongrel log shows that it was issued a restart
command and restarted.
What am I doing wrong? There must be a way to timeout a system level C
API/RFC call…
Running Ruby 1.8.6 p114 on CentOS 5.1 64-bit with mongrel 1.1.5
Thanks in advance for anyone who can help!