[bind10-dev] [sprint planning] ticket discussion for sprint ending 2012-12-04

[JINMEI Tatuya / 神明達哉]

At Mon, 19 Nov 2012 10:52:06 +0100,
Jelte Jansen <jelte at isc.org> wrote:

> There is one spike in the estimate for #2461 (clarify thread
> consideration for DataSourceClient); most estimates are 2 or 3, but
> one is 8. It looks like the ticket is 'just' some internal
> documentation, but either I'm misinterpreting it, or Shane may be

The intent for the ticket is to just clarify the documentation.

> Another big difference is #2468 (b10-stats-httpd_test.py not
> stopping). I believe the source of confusion here is the existing
> discussion in the ticket, and some people think this ticket may have
> been solved already. (FWIW, I think the symptom has only been worked
> around, the source cause has not been addresses, but of course I may
> be wrong)

Right, the subject of the ticket has been changing.  I don't know if
the original problem indicates a problem in stats-httpd or only in the
test, and if it's the former we shouldn't hide it by tweaking the
tests.  But if it's more likely to be the former, my suggestion is:

- create a separate ticket for the "too many open files" issue,
  referring to #2468
- in that ticket, figure out why we needed such a large number of
  sockets after applying the patch suggested in
  http://bind10.isc.org/ticket/2468#comment:2
- and then in that ticket, fix the issue by limiting the creation of
  thread in tests to the case that is really needed (which is
  currently in the trac2468 branch)

We may be able to close #2468 itself.  If the real problem still
exists we should probably need a new one.

---
JINMEI, Tatuya