pids.max cgroup return "max", runc events broken on master

[jessfraz]

If I run runc events on master it just hangs

I git bisected it to 2b1e086

but I cant tell why that broke it...

But on master my cgroups path in /run/runc/container_name/state.json is:

"cgroup_paths": {
    "blkio": "/sys/fs/cgroup/blkio/system.slice/system-runc.slice/chrome",
    "cpu": "/sys/fs/cgroup/cpu,cpuacct/system.slice/system-runc.slice/chrome",
    "cpuacct": "/sys/fs/cgroup/cpu,cpuacct/system.slice/system-runc.slice/chrome",
    "cpuset": "/sys/fs/cgroup/cpuset/system.slice/system-runc.slice/chrome",
    "devices": "/sys/fs/cgroup/devices/system.slice/system-runc.slice/chrome",
    "freezer": "/sys/fs/cgroup/freezer/system.slice/system-runc.slice/chrome",
    "hugetlb": "/sys/fs/cgroup/hugetlb/system.slice/system-runc.slice/chrome",
    "memory": "/sys/fs/cgroup/memory/system.slice/system-runc.slice/chrome",
    "name=systemd": "/sys/fs/cgroup/systemd/system.slice/system-runc.slice/chrome",
    "net_cls": "/sys/fs/cgroup/net_cls,net_prio/system.slice/system-runc.slice/chrome",
    "net_prio": "/sys/fs/cgroup/net_cls,net_prio/system.slice/system-runc.slice/chrome",
    "perf_event": "/sys/fs/cgroup/perf_event/system.slice/system-runc.slice/chrome",
    "pids": "/sys/fs/cgroup/pids/system.slice/system-runc.slice/chrome"
  },

whereas on v0.9 it is correctly:

 "cgroup_paths": {
    "blkio": "/sys/fs/cgroup/blkio/user.slice/htop",
    "cpu": "/sys/fs/cgroup/cpu,cpuacct/user.slice/htop",
    "cpuacct": "/sys/fs/cgroup/cpu,cpuacct/user.slice/htop",
    "cpuset": "/sys/fs/cgroup/cpuset/user.slice/htop",
    "devices": "/sys/fs/cgroup/devices/user.slice/htop",
    "freezer": "/sys/fs/cgroup/freezer/user.slice/htop",
    "hugetlb": "/sys/fs/cgroup/hugetlb/user.slice/htop",
    "memory": "/sys/fs/cgroup/memory/user.slice/htop",
    "name=systemd": "/sys/fs/cgroup/systemd/user.slice/htop",
    "net_cls": "/sys/fs/cgroup/net_cls,net_prio/user.slice/htop",
    "net_prio": "/sys/fs/cgroup/net_cls,net_prio/user.slice/htop",
    "perf_event": "/sys/fs/cgroup/perf_event/user.slice/htop",
    "pids": "/sys/fs/cgroup/pids/user.slice/htop"
  },

[jessfraz]

systemd whyyyyyyyyy you troll me

[runcom]

I get correct results on master:

cat /run/runc/test/state.json  | jq '.cgroup_paths'
{
  "blkio": "/sys/fs/cgroup/blkio/user.slice/test",
  "cpu": "/sys/fs/cgroup/cpu,cpuacct/user.slice/test",
  "cpuacct": "/sys/fs/cgroup/cpu,cpuacct/user.slice/test",
  "cpuset": "/sys/fs/cgroup/cpuset/user.slice/test",
  "devices": "/sys/fs/cgroup/devices/user.slice/test",
  "freezer": "/sys/fs/cgroup/freezer/user.slice/test",
  "hugetlb": "/sys/fs/cgroup/hugetlb/user.slice/test",
  "memory": "/sys/fs/cgroup/memory/user.slice/test",
  "name=systemd": "/sys/fs/cgroup/systemd/user.slice/test",
  "net_cls": "/sys/fs/cgroup/net_cls,net_prio/user.slice/test",
  "net_prio": "/sys/fs/cgroup/net_cls,net_prio/user.slice/test",
  "perf_event": "/sys/fs/cgroup/perf_event/user.slice/test",
  "pids": "/sys/fs/cgroup/pids/user.slice/test"
}

[jessfraz]

ugh what is your systemd version, i have

systemd 229
+PAM +AUDIT +SELINUX +IMA +APPARMOR +SMACK +SYSVINIT +UTMP +LIBCRYPTSETUP +GCRYPT +GNUTLS +ACL +XZ -LZ4 +SECCOMP +BLKID +ELFUTILS +KMOD -IDN

[jessfraz]

ok so its not the json that is the problem, like if i have a container w the systemd.slice json (aka created with master) and use the v0.9 binary to run events for that container it works, but runc events master just hangs

[jessfraz]

this is the weirdest bug ever

[marcosnils]

@jfrazelle I'm about to try this. Let you know in a couple of mins

[jessfraz]

thanks I can't tell if im going insane

[jessfraz]

also i get this when checking the max

$ sudo cat /sys/fs/cgroup/pids/user.slice/htop/pids.max
max

jessie at debian in ~/runc on master
$ sudo cat /sys/fs/cgroup/pids/user.slice/htop/pids.current 
1

[jessfraz]

so if I dont set pids limit in my config I get max if its set i get what i set it to, so i think its a default problem or something

[marcosnils]

@jfrazelle works for me:

marcos@XPS:~/Projects/containers/test$ sudo cat /run/runc/test/state.json  | jq '.cgroup_paths'                       
{
  "blkio": "/sys/fs/cgroup/blkio/user.slice/user-1000.slice/session-c2.scope/test",
  "cpu": "/sys/fs/cgroup/cpu,cpuacct/user.slice/user-1000.slice/session-c2.scope/test",
  "cpuacct": "/sys/fs/cgroup/cpu,cpuacct/user.slice/user-1000.slice/session-c2.scope/test",
  "cpuset": "/sys/fs/cgroup/cpuset/user.slice/user-1000.slice/session-c2.scope/test",
  "devices": "/sys/fs/cgroup/devices/user.slice/user-1000.slice/session-c2.scope/test",
  "freezer": "/sys/fs/cgroup/freezer/user.slice/user-1000.slice/session-c2.scope/test",
  "hugetlb": "/sys/fs/cgroup/hugetlb/user.slice/user-1000.slice/session-c2.scope/test",
  "memory": "/sys/fs/cgroup/memory/user.slice/user-1000.slice/session-c2.scope/test",
  "name=systemd": "/sys/fs/cgroup/systemd/user.slice/user-1000.slice/session-c2.scope/test",
  "net_cls": "/sys/fs/cgroup/net_cls,net_prio/user.slice/user-1000.slice/session-c2.scope/test",
  "net_prio": "/sys/fs/cgroup/net_cls,net_prio/user.slice/user-1000.slice/session-c2.scope/test",
  "perf_event": "/sys/fs/cgroup/perf_event/user.slice/user-1000.slice/session-c2.scope/test"
}

[jessfraz]

you dont have pids cgroup, i figured out the problem its #663 (comment)

[marcosnils]

@jfrazelle good catch. Using old kernels has it's advantages