Skip to content

Latest commit

 

History

History
185 lines (159 loc) · 5.24 KB

cpu.md

File metadata and controls

185 lines (159 loc) · 5.24 KB
Raw
title weight description
Using profile cpu
20
Analyze CPU performance by sampling stack traces.

The profile cpu gadget takes samples of the stack traces.

On Kubernetes

Here we deploy a small demo pod "random":

$ kubectl run --restart=Never --image=busybox random -- sh -c 'cat /dev/urandom > /dev/null'
pod/random created

Using the profile cpu gadget, we can see the list of stack traces. The following command filters only for pods named "random", execute the command and interrupt it after ~30 seconds. The -K option is passed to show only the kernel stack traces.

$ kubectl gadget profile cpu --podname random -K
Capturing stack traces... Hit Ctrl-C to end.^C

After a while press with Ctrl-C to stop trace collection

K8S.NODE         K8S.NAMESPACE    K8S.POD                        K8S.CONTAINER    PID     COMM             COUNT
minikube         default          random                         random           340800  cat              1
        entry_SYSCALL_64_after_hwframe
        do_syscall_64
        __x64_sys_read
        ksys_read
        vfs_read
        urandom_read
        urandom_read_nowarn.isra.0
        extract_crng
        _extract_crng
...
minikube         default          random                         random           340800  cat              9
        entry_SYSCALL_64_after_hwframe
        do_syscall_64
        __x64_sys_read
        vfs_read
        ksys_read
        urandom_read
        urandom_read_nowarn.isra.0
        copy_user_generic_string

From the traces above, you can see that the pod is spending CPU time in the Linux function urandom_read.

Instead of waiting, you can use the --timeout argument:

$ kubectl gadget profile cpu --timeout 5 --podname random -K
Capturing stack traces...
K8S.NODE         K8S.NAMESPACE    K8S.POD                        K8S.CONTAINER    PID     COMM             COUNT
minikube         default          random                         random           340800  cat              1
        entry_SYSCALL_64_after_hwframe
        do_syscall_64
        __x64_sys_read
        ksys_read
        vfs_read
        urandom_read
        urandom_read_nowarn.isra.0
        extract_crng
        _extract_crng
...
minikube         default          random                         random           340800  cat              9
        entry_SYSCALL_64_after_hwframe
        do_syscall_64
        __x64_sys_read
        vfs_read
        ksys_read
        urandom_read
        urandom_read_nowarn.isra.0
        copy_user_generic_string

This gadget also supports custom column outputting, for example:

$ kubectl gadget profile cpu --timeout 1 --podname random -o columns=k8s.node,k8s.pod
Capturing stack traces...
K8S.NODE         K8S.POD
minikube         random
...
minikube         random

The following command is the same as default printing:

$ kubectl gadget profile cpu --timeout 1 --podname random -o columns=k8s.node,k8s.namespace,k8s.pod,k8s.container,pid,comm,count
Capturing stack traces...
K8S.NODE         K8S.NAMESPACE    K8S.POD                        K8S.CONTAINER    PID     COMM             COUNT
minikube         default          random                         random           340800  cat              1
        entry_SYSCALL_64_after_hwframe
        do_syscall_64
        __x64_sys_read
        ksys_read
        urandom_read
        vfs_read
        urandom_read_nowarn.isra.0
        extract_crng
        _extract_crng
        __lock_text_start
        [unknown]
        [unknown]
...
minikube         default          random                         random           340800  cat              1
        entry_SYSCALL_64_after_hwframe
        do_syscall_64
        __x64_sys_read
        ksys_read
        urandom_read
        vfs_read
        urandom_read_nowarn.isra.0
        extract_crng
        _extract_crng
        __lock_text_start
        [unknown]
        [unknown]

Finally, we need to clean up our pod:

$ kubectl delete pod random

With ig

  • Generate some kernel load:
$ docker run -d --rm --name random busybox cat /dev/urandom > /dev/null
  • Start ig:
$ sudo ./ig profile cpu -K --containername random --runtimes docker
  • Observe the results:
sudo ./ig profile cpu -K --containername random --runtimes docker
Capturing stack traces... Hit Ctrl-C to end.^C
RUNTIME.CONTAINERNAME                                                                        COMM             PID        COUNT
random                                                                                       cat              641045     1
        entry_SYSCALL_64_after_hwframe
        do_syscall_64
        __x64_sys_sendfile64
        do_sendfile
        splice_file_to_pipe
        generic_file_splice_read
        get_random_bytes_user
        chacha_block_generic
        chacha_permute
...
random                                                                                       cat              641045     5
        entry_SYSCALL_64_after_hwframe
        do_syscall_64
        __x64_sys_sendfile64
        do_sendfile
        splice_file_to_pipe
        generic_file_splice_read
        get_random_bytes_user
        chacha_block_generic
        chacha_permute
  • Remove the docker container:
$ docker stop random