kworker at 100% CPU - find_get_pages_tag

[thevsdr]

Hey guys,

I have just installed ubuntu-mate on 17.04 on my Dell Precision 5510, it has 16GB of ram, i7-6820HQ [running intel-microcode] and a Quadro M1000M [ running nvidia-375]. Kernel version is: 4.10.0-21-generic.

I initially installed this without a swap as 16gb of ram "should" be ok for what i needed however apps started to crash due to "out of memory", i checked "free -m" showed it was all cached but still ~10gb available..
This is odd, but figured i underestimated the need for swap in this case so i added a 32GB swap partition and now i have kworker/u16:4 running at 100% with the the below trace:


Samples: 156K of event 'cycles', Event count (approx.): 105130510564Overhead Command Shared Object Symbol ◆
50.43% kworker/u16:4 [kernel.kallsyms] [k] find_get_pages_tag ▒
19.30% kworker/u16:4 [kernel.kallsyms] [k] write_cache_pages ▒
12.32% kworker/u16:4 [kernel.kallsyms] [k] release_pages ▒
10.32% kworker/u16:4 [kernel.kallsyms] [k] unlock_page ▒
5.87% kworker/u16:4 [kernel.kallsyms] [k] radix_tree_next_chunk ▒
0.65% kworker/u16:4 [kernel.kallsyms] [k] _cond_resched ▒
0.35% kworker/u16:4 [kernel.kallsyms] [k] lru_add_drain_cpu ▒
0.23% kworker/u16:4 [kernel.kallsyms] [k] pagevec_lookup_tag ▒
0.14% kworker/u16:4 [kernel.kallsyms] [k] __pagevec_release ▒
0.08% kworker/u16:4 [kernel.kallsyms] [k] mem_cgroup_uncharge_list ▒
0.07% kworker/u16:4 [kernel.kallsyms] [k] free_hot_cold_page_list ▒
0.02% kworker/u16:4 [kernel.kallsyms] [k] _aesni_enc1 ▒
0.01% kworker/u16:4 [kernel.kallsyms] [k] vma_interval_tree_subtree_search ▒
0.01% kworker/u16:4 [kernel.kallsyms] [k] update_blocked_averages ▒
0.01% kworker/u16:4 [kernel.kallsyms] [k] smp_call_function_many ▒
0.01% kworker/u16:4 [kernel.kallsyms] [k] rcu_check_callbacks ▒
0.01% kworker/u16:4 [kernel.kallsyms] [k] intel_pstate_update_util ▒
0.01% kworker/u16:4 [kernel.kallsyms] [k] perf_event_task_tick ▒
0.01% kworker/u16:4 [kernel.kallsyms] [k] _raw_spin_lock ▒
0.01% kworker/u16:4 [kernel.kallsyms] [k] update_wall_time ▒
0.01% kworker/u16:4 [kernel.kallsyms] [k] native_queued_spin_lock_slowpath ▒
0.01% kworker/u16:4 [kernel.kallsyms] [k] cpuacct_charge ▒
0.00% kworker/u16:4 [kernel.kallsyms] [k] update_load_avg ▒
0.00% kworker/u16:4 [kernel.kallsyms] [k] irq_exit

This looks to me that the OS is having problems with the swap and paging in and out however free shows there is only 937mb in swap and plenty of "Available" memory

$ free -m total used free shared buff/cache available
Mem: 15860 2495 150 170 13214 12851
Swap: 31999 937 31062




The only other thing of note, i am also running a windows VM in VMware Workstation which is the biggest ram hog set at 4096 MB and 1 CPU, 2 Cores. Intel VT-x/EPT. This should not be causing this and the vmware-vmx process is not showing anything abnormal.

PID TID RUID EUID THR SYSCPU USRCPU VGROW RGROW RDDSK WRDSK ST EXC S CPUNR CPU CMD 1/2
6304 - root root 1 9.99s 0.00s 0K 0K 0K 7584K -- - R 2 100% kworker/u16:4
4286 - wardd wardd 23 6.14s 0.06s 0K -20.3M 32K 9204K -- - S 0 62% vmware-vmx
9391 - wardd wardd 16 0.10s 0.93s 0K 572K 0K 0K -- - S 5 10% chrome
2638 - wardd wardd 5 0.08s 0.49s 1280K 1148K 0K 0K -- - S 0 6% chrome





Anyone know what i can do to either force the kernel to release this ram or some how restrict how much it is caching?

Thank you