BIND 10 #1314: IXFR-out system tests

[BIND 10 Development]

#1314: IXFR-out system tests
-------------------------------------+-------------------------------------
                   Reporter:         |                 Owner:  jinmei
  stephen                            |                Status:  reviewing
                       Type:  task   |             Milestone:
                   Priority:  major  |  Sprint-20111220
                  Component:         |            Resolution:
  Unclassified                       |             Sensitive:  0
                   Keywords:         |           Sub-Project:  DNS
            Defect Severity:  N/A    |  Estimated Difficulty:  7
Feature Depending on Ticket:  IXFR-  |           Total Hours:  5
  out                                |
        Add Hours to Ticket:         |
                  Internal?:  0      |
-------------------------------------+-------------------------------------
Changes (by jelte):

 * owner:  jelte => jinmei


Comment:

 Replying to [comment:14 jinmei]:
 >
 > Yeah, and I think we need more generic and reliable way to confirm
 > that a component is fully ready, not only for tests.  For example, I
 > guess b10-auth would like to know whether xfrout/ddns is really active
 > rather than just naively passing requests to the corresponding UNIX
 > domain socket (and possibly getting an exception subsequently).
 >
 > One thing we could do for tests would be to implement the 'ping'
 > command for all components.
 >

 Yes. One problem with that (at least in this context), is that it is, how
 do you say it, a pull-type request (you need to ping it, if it has not
 started, you need to ping it again after some delay). The nice thing
 (again, in this context) about the current approach, if supported through
 log messages, it that you just wait until the message comes by. Originally
 I had thought of having some sort of announcement channel in the msgq
 where modules would shout "Hi guys, I'm up!", but such things have been
 disappeared in the backlog somewhere :p. Either approach would be very
 good to have though.

 > Anyway, that will be beyond the scope of this ticket.
 >

 ack.

 >
 > Yes, it now succeeds constantly.
 >
 > "NO_ANSWER" and "timeout" seem to suggest the same thing might be
 > happening?  (again, would be a separate issue though).
 >

 AFAIK every test should have something similar, especially for queries,
 but I'd need to take a look at the specific test that fails and perhaps
 its output. but not for this ticket indeed :)

 >
 > I've now quickly looked at others a bit more closely.  Some comments:
 >
 > - What's the difference between Scenario 1's tests 1 and 2?

 Oops, serial of test two should be 22. Changed.

 Also just noticed I made a similar mistake in scenario 2 test 3.


 > - I guess many of the tests can be done successfully over TCP and
 >   wonder we should actually do it?

 Some of them can (in fact, I wrote those using tcp in the first place). I
 reverted them before submitting to match the list on the IxfrSystemTest
 page, but yeah, we can use TCP.

 For those where it is possible, I changed udp to tcp, with a comment that
 the original requirement is for udp. Added a note at the top that we might
 want to keep the tcp tests once we have udp support, and perhaps reflect
 that back on the systest page then too.

 For all tests I either added a comment that they are changed to TCP for
 now or that they are commented out because they are not supported yet and
 have no TCP equivalent.

 Note: I did not do the first one, we actually fail here (I hope because
 this branch does not do serial comparison yet)

-- 
Ticket URL: <http://bind10.isc.org/ticket/1314#comment:16>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development