BIND 10 #324: improve SQL data source performance

[BIND 10 Development]

#324: improve SQL data source performance
-------------------------------------+-------------------------------------
                   Reporter:  each   |                 Owner:  jinmei
                       Type:         |                Status:  reviewing
  enhancement                        |             Milestone:
                   Priority:         |  Sprint-20120417
  medium                             |            Resolution:
                  Component:  data   |             Sensitive:  0
  source                             |           Sub-Project:  DNS
                   Keywords:         |  Estimated Difficulty:  5
  performance                        |           Total Hours:  0
            Defect Severity:  N/A    |
Feature Depending on Ticket:         |
  sqlite schema upgrade              |
        Add Hours to Ticket:  0      |
                  Internal?:  0      |
-------------------------------------+-------------------------------------

Comment (by jinmei):

 Replying to [comment:25 jelte]:

 > > That said, I tried to fix it in the branch anyway (4df83ec).  But I
 > > couldn't come up with a reasonable test scenario, so I didn't add any
 > > test.  So, I'm okay either with or without incorporating this
 > > particular change to master.
 >
 > this looks fine and I propose we include it. One trivial change could be
 to add Exclusive to the class name so we don't accidentally use it for
 normal transactions in the future. But i'll leave that up to you :)

 Okay.  I'll keep the class name simply because I believe we should
 stop the DB creation here pretty soon.

 > > As for newschema.sqlite3, yes, that's a good point.  I changed the
 > > version to 1000.  Regarding the need for updating test DB files in
 > > general, I think it's better to create test data at build/run time
 > > dynamically (except for the intentionally unusual/broken ones).  But
 > > I believe it's beyond the scope of this task and should go to a
 > > separate ticket (maybe one of "hardening" things).
 >
 > ok. Speaking of which, I found one other problem; the zonemgr tests have
 the side-effect of doing just this, by creating a database called
 initdb.file, and leaving it (until you run make clean). So if you switch
 to and from this branch (or later, if you pull the merge of this into
 master) and don't run make clean, these tests will fail with a database
 error.
 >
 > AFAICT, this file does not need to be kept, and we might as well delete
 it at the start of make check, i.e. something like

 Hmm.  While this would solve this particular situation, more
 fundamentally I think this file must be always cleaned up after each
 test case (and it probably has to be assured that it doesn't exist at
 the beginning of each test case).  Further, ideally I guess we should
 even avoid creating such a file with help of some abstraction.

 I revised the branch with the first type of cleanup.  I think The
 second one is beyond the scope of this ticket.  The change is not
 trivial, so I'll give the ticket back to you once again.

-- 
Ticket URL: <http://bind10.isc.org/ticket/324#comment:26>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development