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

[cyphar]

@jfrazelle Sorry, this was my fault. There's a fix in #644. Basically, it's because I forgot to account for the string "max" so parsing fails. The reason it wasn't merged sooner is because there was some discussion about what value we should return to represent "max".

[jessfraz]

oh nice i had opened one too :) but your will hopefully get in first!

I closed mine :)

[mrunalp]

@jfrazelle Can we close this now that #644 is merged? :)

[jessfraz]

thanks for reminding me sorry

[mrunalp]

Thanks! No worries :)