Interview questions – one approach
Posted by John Hallas on March 8, 2012
In my last post I talked about doing a lot of interview questions lately. I have been thinking of changing my standard set of questions anyway so I thought it would be a good idea to write about the questions I have been asking and the responses I get.
Firstly let me see that the question itself is not important, what is more important is where it leads me and does it help me find out more about the candidate than I knew previously. Secondly, there is no right or wrong answer. If the interviewee does not have a clue but makes an educated guess based on other knowledge that is much better than just saying ‘I don’t know’, or what is more common, humming and aarring for 5 minutes hoping the answer will pop into his/her head. PS I think aarring is a genuine word but I don’t think it is spelt like that.
Someone else normally asks about other areas such as Dataguard, RAC, builds, RMAN, patching etc. Not because I am not interested or aware but so that we get two different viewpoints. I tend to loosely ask about ‘performance’. I do not think I have seen a CV recently that does not claim to have tuned a database to perfection. Normally on further inquiry it breaks down to making the redo logs 500Mb instead of 100K or something similar. I start off asking about a 10046 trace and what values you can provide to it to give different output. If that goes OK I ask about what a 10053 trace gives you (optimizer choices looked at and chosen)).
Adaptive cursor sharing normally gets a poor response but bind variable peaking often triggers a response
The effect of the init.ora parameter optimizer_dynamic _sampling provides some interesting responses and then moving on to when stale stats and no stats on tables and how the optimizer responds allows me to delve deeper.
If you see the wait event log_file_sync then what areas would you investigate? Everyone says resize redo logs but I am looking for commit frequency as well as disk access speeds. Someone recently just answered with “disk I/O” which was quite a relief after all the prompting that is normally required.
Another favourite is “are there any downsides to using a RMAN catalogue” which tends to stump people as they are not used to looking at the negatives. I am thinking around the requirement for another server/database and the licensing/hw costs and also the problem that if the catalogue cannot be accessed then a backup will fail. Again the common belief is that a RMAN catalogue should be used but people struggle to identify exactly why.
I have only loosely covered the points I talk about above and don’t pretend that I have covered all the angles possible but if someone talks through competently about all those then I feel we are on good ground.
I will now go off and think of some more questions to add to my armoury and if anyone if anyone is looking for permanent employment in the North of England we do have vacancies. I can be found on linkedin.
PS I sometimes ask what blogs and oracle material the candidate uses. I normally have a grin on my face at that point which is a bit of hint as to what answer I am looking for. I don’t mind Jonathan Lewis or Tom Kyte being mentioned before me but I draw the line after that.