Multiple concurrent request to Docker can cause it to get unresponsible

[gomezgoiri]

Related issue: docker/docker-py#398

[gomezgoiri]

I have tried to use pause/unpause commands to see if a paused container consume less CPUs (although they will still consume memory).

"...when you put a process in the freezer (or its cgroup), you block the execution of new task
 of that process from the kernel task scheduler (i.e.: it stops the process), but you don't kill
them and they keep consuming the memory they are using (although the Kernel may move
that memory to swap or to solid disk). And the CPU resources used by a paused container
I would consider insignificant."

From the response time point of view, after unpausing it gives the same performance as any other already started Packet Tracer instance.

While docker stats says that it consumes 0%, CAdvisor says it keeps consuming memory.

After checking the consumption using ps -ax | grep "PacketTracer6", I noticed that the time associated doesn't increase while the container is paused.

(benchmark)[agg96@kmi-appsvr05 ptdockertest]$ ps -ax | grep "PacketTracer6"
112677 ?        S      0:00 /bin/bash -c unset http_proxy; /opt/pt/bin/PacketTracer6 saves/FORGE/ibookdemo611.pkt
112678 ?        Sl     1:08 /opt/pt/bin/PacketTracer6 saves/FORGE/ibookdemo611.pkt
113206 pts/0    S+     0:00 grep --color=auto PacketTracer6

[gomezgoiri]

Another way to avoid blocking Docker could be to reserve one core not to be used by Packet Tracer containers.

For this, we could use docker run --cpuset="1-15" making sure that we don't block core 0.