help interpret flame graph

[Scott Guyer]

Hi folks,

  We have an application that we are profiling. It's not getting anywhere near the throughput we would expect. As I look at this flame graph, it seems to be spending the bulk of the time in LuaJIT and lpeg (lua parser).  This doesn't match expectations because (1) the data was collected over a 5 minute load test; (2) we have the lua_code_cache on. 

So I'm thinking we either screwed something up in the config, or I'm completely misreading the flame graph.

I appreciate any tips this community may offer. Thanks!

-Scott

NOTE: perf data was captured using ...

perf record -F 99 -a -g -- sleep 300

perf script | ./stackcollapse-perf.pl > out.perf-folded

./flamegraph.pl out.perf-folded > perf-kernel-nologging.svg

[Robert Paprocki]

Scott,

There's no useful data in the flame graph. For profiling specifics about your luajit calls it's probably best to use the system tap extensions for openresty:

https://github.com/openresty/stapxx

Being able to drill into the Lua VM is necessary to profile your application beyond indicating "yep it's Lua alright!" (which, as you've observed, is less than helpful ;))

Good luck with it!

--
You received this message because you are subscribed to the Google Groups "openresty-en" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openresty-en...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

<perf-kernel-nologging.svg>