diff options
author | Giuseppe Scrivano <gscrivan@redhat.com> | 2021-09-09 17:16:24 +0200 |
---|---|---|
committer | Giuseppe Scrivano <gscrivan@redhat.com> | 2021-09-10 09:03:10 +0200 |
commit | 50688da29b4a37006d0a7fafc25aca1a6447c7d9 (patch) | |
tree | 18c3d173262d247c8d505a22952ba238942e313e | |
parent | 2b85382ca187fdcc000803b1d26d8b46e8cb755f (diff) | |
download | podman-50688da29b4a37006d0a7fafc25aca1a6447c7d9.tar.gz podman-50688da29b4a37006d0a7fafc25aca1a6447c7d9.tar.bz2 podman-50688da29b4a37006d0a7fafc25aca1a6447c7d9.zip |
stats: detect containers restart
if the current cpu usage time is lower than what previously recorded,
then it means the container was restarted and now it runs in a new
cgroup. When this happens, reset the prevStats.
Closes: https://github.com/containers/podman/issues/11469
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
-rw-r--r-- | libpod/stats.go | 6 |
1 files changed, 6 insertions, 0 deletions
diff --git a/libpod/stats.go b/libpod/stats.go index 2532b35c2..975152535 100644 --- a/libpod/stats.go +++ b/libpod/stats.go @@ -54,6 +54,12 @@ func (c *Container) GetContainerStats(previousStats *define.ContainerStats) (*de return nil, err } + // If the current total usage in the cgroup is less than what was previously + // recorded then it means the container was restarted and runs in a new cgroup + if previousStats.Duration > cgroupStats.CPU.Usage.Total { + previousStats = &define.ContainerStats{} + } + previousCPU := previousStats.CPUNano now := uint64(time.Now().UnixNano()) stats.Duration = cgroupStats.CPU.Usage.Total |