Hi,
I'm running Rack application on uWSGI (1.3-rc3) on CentOS5. While
testing reliability of Zerg dance on one end I run client making
requests and on the other end I was restarting uwsgi in graceful manner
in a loop.
For most of the time it works just fine, but occasionally I get HTTP 504
on the client. Stats server shows one worker in a busy state whereas
other workers are in pause. Strace on busy worker gives:
23568 12:53:42.054778 clock_gettime(CLOCK_REALTIME, {1348224822, 54804000}) = 0
<0.000011>
23568 12:53:42.054879 pipe([6, 8]) = 0 <0.000022>
23568 12:53:42.054969 pipe([9, 10]) = 0 <0.000014>
23568 12:53:42.055033 fcntl(10, F_SETFD, FD_CLOEXEC) = 0 <0.000011>
23568 12:53:42.055092 rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0 <0.000010>
23568 12:53:42.055166 futex(0x2ab153ea4e00, FUTEX_WAIT_PRIVATE, 2, NULL) = -1
EINTR (Interrupted system call) <3.565238>
23568 12:53:45.620466 --- SIGTSTP (Stopped) @ 0 (0) ---
23568 12:53:45.620506 rt_sigreturn(0x14) = -1 EINTR (Interrupted system call)
<0.000009>
23568 12:53:45.620556 futex(0x2ab153ea4e00, FUTEX_WAIT_PRIVATE, 2, NULL
Worker stays in busy state forever.
Increasing uwsgi_read_timeout in nginx does not help.
Not sure if it's an app or uWSGI problem..
Cheers,
Marcin
_______________________________________________
uWSGI mailing list
[email protected]
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
