On Wed, Sep 28, 2011 at 3:35 PM, Brian G= effon <briang= effon@gmail.com> wrote:

Hi,
I'm encountering something strange with ATS3.0.1 on Red Hat Ente= rprise Linux 6, using a vanilla build with no modules enabled and the defau= lt records.config /w zero entries in the remap file, ATS is idling at very = high CPU (around 15-20%).

root =A0 =A0 13509 =A00.0 =A00.0 =A058512 =A02392 ? =A0 =A0 =A0 =A0Ss = =A0 19:15 =A0 0:00 /usr/local/ats3.0.1plain//bin/traffic_cop
nobody =A0 = 13511 =A00.1 =A00.0 480072 16632 ? =A0 =A0 =A0 =A0Sl =A0 19:15 =A0 0:00 /us= r/local/ats3.0.1plain/bin/traffic_manager
nobody =A0 13521 ***16.9*** =A00.1 1584628 114224 ? =A0 =A0 =A0Sl =A0 19:15= =A0 2:06 /usr/local/ats3.0.1plain/bin/traffic_server -M -A,7:X
So I used strace to try to determine what might be causing this= , and here is what i've found:

[root@machine]# strace -c -p 13521
Proce= ss 13521 attached - interrupt to quit
^CProcess 13521 detached
% time =A0 =A0 seconds =A0usecs/call =A0 =A0 calls =A0 =A0errors sy= scall

------ ----------- ----------- --------- --------- ----------------
100.00 =A0 =A03.589451 =A0 =A0 =A0 =A0 796 =A0 =A0 =A04510 =A0 =A0 = =A0 =A0 =A0 epoll_wait
------ ----------- ----------- --------- -= -------- ----------------

100.00 =A0 =A03.589451 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A04510 =A0 =A0= =A0 =A0 =A0 total
[root@machine]#=A0

<= div>
It appears that it's entirely epoll_wait, and each c= all is taking 796 microseconds! So I have to concerns with this, first, why= would epoll_wait take such a long amount of time, 796 microseconds seems l= ike a long time, and more importantly, how could it possibly be called so f= requently, does ATS use a short timeout when doing epoll_waits?

I would really=A0appreciate=A0any feedback regarding th= is, has anyone else experienced this? Is there anywhere else I might look t= o determine the cause of this? Could this be classified as _normal_=A0behav= ior?

It is workin= g as designed. The epoll timeout is set to 0 or 10 msec on linux. Could the= re be very little or no load going through this instance? That would explai= n why you are only seeing epoll_wait. In that case there isn't really a= nything wrong, the process is just burning through epoll_waits looking for = something to do when nothing is available. Once the process starts taking e= nough traffic, you will start to see user space and other kernel functions = start to take cpu time.

=A0Sridhar