os error 14 on Ubuntu about rbspy HOT 9 CLOSED

Thanks for reporting this! I tried it on my computer and got this flamegraph, so I don't think it's to do with the Ruby code here.

How did you install Ruby? It would help if you could send me a core dump or two of the Ruby process's memory so I can see if I can figure out what's going on. Here's how to get a core dump:

first, start a while loop that will try to grab a core dump:

while true
do 
sudo gcore $(pgrep -f jekyll)
sleep 0.5
done

and then run bundle exec jekyll build

gcore will generate a file named like core.24996, which is the core dump.

you can either upload them somewhere or email them to [email protected].

from rbspy.

I'm using ruby through rvm. Here is the output of ruby -v:

ruby 2.4.2p198 (2017-09-14 revision 59899) [x86_64-linux]

Here is the incomplete flamegraph I initially got:

And here are two dumps:

• http://tmp.pixelastic.com/share/core.19844
• http://tmp.pixelastic.com/share/core.21121

Thanks!

from rbspy.

I am getting a similar error message when I run on EC2 using Amazon Linux AMI 2018.03.0.20180508 x86_64 HVM

https://homepage.cs.uiowa.edu/~hawks/core.16680.zip

from rbspy.

I tried several different operating systems and machines, but could not get rbspy to work with RVM.

If I instead use rbenv it works without issue.

from rbspy.

It looks like what's happening here is that RVM has some kind of hooking logic that launches a Ruby process running ruby_executable_hooks, then execv's Bundler, wiping out the address space and starting up a new instance of the interpreter. The memory address that cannot be copied is the pointer to the initial Ruby thread.

The easy workaround here is that it seems RVM does all this indirection in service of allowing rbspy record -- jekyll build, which works. But more generally it would be nice if rbspy could handle execv calls. (I could imagine this coming up in programs using fork-then-execv strategies too.) I've hacked together some code that takes the naive approach of reinitializing the StackTraceGetter if the pointer to stack becomes invalid and it seems to work. Would a PR along those lines be of interest?

from rbspy.

yes, a PR along those lines sounds great. Thanks for investigating this! I'd love to get rbspy working with RVM.

from rbspy.

I'm getting a similar error attaching rbspy to a rake task without RVM:

[mymachine]:~$ sudo RAILS_ENV=production rbspy record bundle exec rake profile:db_size
Dropping permissions: running Ruby command as user ubuntu
Press Ctrl+C to stop
Dropped 21/33 stack traces because of errors. Please consider reporting a GitHub issue -- this isn't normal.
Wrote raw data to /home/ubuntu/.cache/rbspy/records/rbspy-2019-01-30-IxHGf8Be38.raw.gz
Writing formatted output to /home/ubuntu/.cache/rbspy/records/rbspy-2019-01-30-01i0vZzsrM.flamegraph.svg
Error. Causes: 
- Failed to copy memory address 7f6405991320
- Bad address (os error 14)

The ruby version is 2.2.7 coming from ubuntu package

from rbspy.

@pixelastic #177 should have fixed your issue. Can you confirm this is the case?

from rbspy.

Thanks for taking the time to investigate and fix this. Unfortunately I moved to other projects, stacks and environments in the meantime and can't try the fix and confirm if it's working.

I'm still going to close this, I'll re-open if I ever encounter this issue again.

Thanks for all your work

from rbspy.