Monday March 28, 2005

Why do I like logging bugs ?

Stacey wrote a good article on how customers can request changes to Solaris. I am pleased to say I have written a few RFE's that have gone into changing Solaris. I logged a very simple bug for the manual page for the Sunray command utadm last week. The -f option is meant to offline the server but leave existing sessions intact. For the 3.0 release the page reads : Existing sessions are killed, but load balancing does not select this server for new sessions.. The word 'not' have been removed. Simple typo. A simple bug but for new Sunray customers this could cause confusion or misunderstanding. At the moment I seem to have two other active bugs where I am in communication with Engineering. I am kind of annoyed with myself on the Sunray bug and a few others we have logged recently because if we had not been moving the lab we would have found time to run the beta products and found these bugs before FCS... But we are doing our bit now.

Empowering, encouraging and rewarding people is what we try in Sun to do to make sure people do log bugs when they find problems. It is a bit like the story of a village which built a fountain and wanted milk to run in it. Everyone was told to put a pint of milk in the fountain. Some folks said "no one will notice if I put water in the fountain instead". Of course they turned the fountain on and it ran with clear water.

So when I hear people say this core dumped or this did not work I tell them to log a bug. Simply running pstack on a core file can often be used to assertain if the problem is a known issue by putting the stack trace into Sunsolve. If it is not then I would log a bug giving as much detail as possible and attaching relevant message files, patch lists and so on. Sometime you have to put a lot of effort into helping the responsible engineer (RE) gather more data. I view this as a learning exercise the process usually goes log a bug and the RE then contacts you for more information. It can be a very rewarding experience, watching the trail of information requests and the final bug fix.

I also wonder if in the future customers would pay for a service where all there 'system' program core dumps were analysed for them. You would use coreadm to setup a global cores area ( note this has changed in Solaris10 to include a zone name in the filename for the core file) and then something would spot new core files, check to see if they were known issues, email the customer advising to load a new patch,that a new bug is now logged or there is and existing bug etc etc.