[Gluster-devel] Script to identify ref leaks

[Niels de Vos]

On Fri, Mar 31, 2017 at 02:34:04PM +0530, Sanoj Unnikrishnan wrote:
> +Mohit
> Mohit had written a similar script to RCA an issue couple of months back.
> 
> It would help if we placed the scripts, tapsets in the source tree itself
> (maybe in a directory under glusterfs/extras/).
> We could also have the tapset packaged into debuginfo packages and deployed
> under /usr/share/systemtap/ path upon installation.

Yes, or we can place them in their own repository. I have a
gluster-debug repository where I plan to put tools I use for debugging.
This repository may well live at the Gluster organization on GitHub too.

https://github.com/nixpanic/gluster-debug

Many of the tools we use for debugging should not be version specific,
so having them in the glusterfs repository might be a little awkward?

What are the opinions of others?

Thanks,
Niels


> Regards,
> Sanoj
> 
> On Fri, Mar 31, 2017 at 12:44 PM, Sonal Arora <sarora at redhat.com> wrote:
> 
> > Hi,
> >
> > I am working on finding ways to indentify ref leaks in glusterfs.
> >
> > Description of Issue : https://bugzilla.redhat.com/show_bug.cgi?id=1417539
> >
> > Goal : To make script which could detect all kind of ref leaks.
> >
> > Script : https://github.com/SonaArora/Tracing-userspace-app/blob/
> > master/ref-leak/try8-modified
> > The above script is a POC program to depict the idea of how to identify
> > leaks. Script is probing dict_ref() and dict_unref() and keeping a track of
> > the pointers, back traces which are referenced/dereferenced by above
> > functions. If the count of refs is unequal to unrefs for each pointer,it
> > will print all the traces corresponding to the leaked pointer.
> > Output : https://github.com/SonaArora/Tracing-userspace-
> > app/blob/master/ref-leak/output-refleak
> > I am working on post processing the output - to filter only the leaked
> > traces and to write the output after every few hours into a file.
> > The script can be extended to all objects being referenced (like
> > inodes/fds).
> >
> > End Goal : Future goal is to make it more versatile, dynamic and light
> > weight so that it can be even utilized on production environments without
> > utilizing much of the system resources.
> >
> > Request your comments and suggestions.
> >
> > Best
> > -Sonal
> >
> >
> >
> > _______________________________________________
> > Gluster-devel mailing list
> > Gluster-devel at gluster.org
> > http://lists.gluster.org/mailman/listinfo/gluster-devel
> >
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: not available
URL: <http://lists.gluster.org/pipermail/gluster-devel/attachments/20170331/92af4d4e/attachment.sig>