[Solved] Lighttpd FastCGI process in tight ioctl(FIONREAD) spin loop in lighttpd 1.4.49
Added by LiveFreeAndRoam@gmail.com almost 5 years ago
Lighttpd v1.4.49.
The lighttpd process (on rare occassions) stops responding to requests. From strace, I observed it is in a tight spin loop:
$ top Mem: 523408K used, 215020K free, 0K shrd, 538163K buff, 4294967295K cached CPU: 13.9% usr 51.0% sys 0.0% nic 35.0% idle 0.0% io 0.0% irq 0.1% sirq Load average: 2.03 1.82 1.47 3/453 10616 PID PPID USER STAT VSZ %VSZ CPU %CPU COMMAND 4830 4826 sysadmin R 6284 0.8 1 46.9 /usr/sbin/lighttpd -D -f /conf/lighttpd.conf.14.49
$ strace -p 4830 Process 4830 attached ioctl(26, FIONREAD, [4922932]) = 0 ioctl(26, FIONREAD, [4922932]) = 0 ioctl(26, FIONREAD, [4922932]) = 0 : ... at infinitum ... :
It is not possible to gracefully kill this process, so instead I use: "kill -9 4830".
I scanned the release notes 1.4.50-1.4.54, but nothing stands out as addressing this. Since this condition seldom occurs, but is devastating when it does occur, what else can I do to mitigate it, or get closer to the root cause?
Thanks,
Matt.
Replies (1)
RE: Lighttpd process in tight ioctl(FIONREAD) spin loop - Added by LiveFreeAndRoam@gmail.com almost 5 years ago
After some more searching, I found a similar report:
[Solved] Lighttpd Worker process stuck, CPU usage at 100% - https://redmine.lighttpd.net/boards/2/topics/8091?r=8099#message-8099
This suggests the problem was fixed in 1.4.50.
Matt.