oomd breaks in case of missing memory controller about oomd

[danobi]

Is this reproducible? Also, is there a delay in time between those two log blocks?

If no on the first and yes on the second question, it might be a race related to #25 . I'll put up a fix for that first and see if it helps.

[disaster123]

No, there's no delay. Yes it happens always if action kill_by_memory_size_or_growth is triggered. No it's not a race that cgroup did never exist.

[danobi]

Ah that makes sense. I think I know the issue. Putting up a fix shortly.

We've got someone addressing the root issue -- oomd plugins should be receiving all their information from core oomd. Right now a lot of plugins independently pull cgroup information from the filesystem. This is bad b/c we'd rather not have every plugin reimplement error checking. It also makes tests harder to write.

[danobi]

Please reopen if 248ff9f didn't fix the issue.

[disaster123]

No idea how to reopen - but no it does not fix the problem.

Output is now:

[2019-06-23 10:35:37] [../util/Fs.cpp:162] Unable to open /sys/fs/cgroup/systemultralowprio.slice/clamav-freshclam.service/memory.current
[2019-06-23 10:35:37] [../Oomd.cpp:527] Caught exception: /sys/fs/cgroup/systemultralowprio.slice/clamav-freshclam.service: missing file

crash / exit of oomd

[disaster123]

@danobi please reopen

[danobi]

Sorry about the delay. Can you post your oomd config? It might be another plugin that's pulling cgroup fs data independently.

[disaster123]

sure:

{
    "rulesets": [
        {
            "name": "mem protection",
            "detectors": [
                [
                    "mem pressure above 10 for 10s",
                    {
                        "name": "dump_cgroup_overview",
                        "args": {
                            "cgroup": "user.slice/*.service,system.slice/*.service,systemlowprio.slice/*.service,systemultralowprio.slice/*.service,user.slice/*.slice,system.slice/*.slice,systemlowprio.slice/*.slice,systemultralowprio.slice/*.slice",
                            "always": "false"
                        }
                    },
                    {
                        "name": "pressure_above",
                        "args": {
                            "cgroup": "user.slice/*.service,system.slice/*.service,systemlowprio.slice/*.service,systemultralowprio.slice/*.service,user.slice/*.slice,system.slice/*.slice,systemlowprio.slice/*.slice,systemultralowprio.slice/*.slice",
                            "resource": "memory",
                            "threshold": "10",
                            "duration": "10"
                        }
                    }
                ]
            ],
            "actions": [
                {
                    "name": "kill_by_memory_size_or_growth",
                    "args": {
                            "cgroup": "user.slice/*.service,system.slice/*.service,systemlowprio.slice/*.service,systemultralowprio.slice/*.service,user.slice/*.slice,system.slice/*.slice,systemlowprio.slice/*.slice,systemultralowprio.slice/*.slice",
                         "post_action_delay": 30,
                         "size_threshold": 95
                    }
                }
            ]
        }
    ],
    "version": "1.0.0"
}

[danobi]

Thanks.

#73 should fix it. It might be a day or two before it's merged -- the usual code review delays. If you get a chance, could you try out the patch?