This issue is caused by high frequency socket polling (at 10 ms intervals). I have created a patch that introduces an adaptive timeout for socket polling, which eliminates this issue. See more at https://github.com/kr/beanstalkd/pull/169
On Tuesday, February 12, 2013 8:28:25 AM UTC+1, Jacob wrote: > > I experience that beanstalkd uses a lot CPU cycles when the system is > under high load. This, even when beanstalkd is not in use. > > This simple test shows the issue in effect: > * beanstalkd v1.8 on an Amazon EC2 Large instance with Ubuntu Linux > 12.04.2 LTS. > * No jobs in beanstalkd > * Saturate the available CPU cycles with a couple of nice'd dummy > processes, using: > `nice -n 10 python -c "while True: x=2**10"&` > > In this test beanstalkd (according to top) uses between 15% and 20% of the > CPU cycles. As soon as the high load is stopped, beanstalkd goes back to > nearly ~0% CPU. > > This observed behaviour is quite unfortunate. So what is happening, and is > it expected behaviour? > -- You received this message because you are subscribed to the Google Groups "beanstalk-talk" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/beanstalk-talk?hl=en. For more options, visit https://groups.google.com/groups/opt_out.
