[Linux Bash Commands] Basic Bash commands #linux

bash_commands.sh

# Process and Thread commands
strace -p 7577 -f
ps -eLf | grep gl2k | grep 13301
strace -p 13301


perf top
perf record -F 99 -ag -p <pid> -- sleep 10
perf script


DEMO
# PID 1306 MAIN
# PID 12564 - Child that may hang


GL2K

1303
Child PID 9683



# NUMA
# https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/7/html/virtualization_tuning_and_optimization_guide/sect-virtualization_tuning_optimization_guide-numa-auto_numa_balancing

# https://www.thegeekdiary.com/understanding-automatic-numa-balancing/


# TRANSPARENT HUGE PAGES
# https://access.redhat.com/solutions/46111


# strace -p 14603

Process 14603 attached - interrupt to quit

futex(0x7f5c8e6019d0, FUTEX_WAIT, 14604, NULL

^C <unfinished ...>

Process 14603 detached

-> no more output


# strace -F -p 14603 

Process 14603 attached with 8 threads - interrupt to quit

-> + output from all the threads in the process.



As far as I can tell, your bug appears to be 100% cpu on 1 cpu core. This bug is more characterised by less cpu burn as threads sleep when they should be woken ("…thus preventing tasks sleeping forever if wakers don't acknowledge all possible waiters" ), so I would assume that your issue is different (as a starting point) and try to gather more evidence to prove it either way.

run 

sudo perf record -F 99 -ag -p <pid> -- sleep 10

sudo perf script 



To get some more info on the stacks associated with the cpu burn. The stacks may be corrupted by the jvm preventing perf from reporting something useful with -g,  so YMMV, to fix that it's a custom build of openjdk...

Thanks,