BIND 10 #1613: auth per rcode statistics
BIND 10 Development
do-not-reply at isc.org
Thu Feb 16 14:53:12 UTC 2012
#1613: auth per rcode statistics
-------------------------------------+-------------------------------------
Reporter: | Owner: jinmei
jinmei | Status: reviewing
Type: task | Milestone:
Priority: major | Sprint-20120221
Component: | Resolution:
statistics | Sensitive: 0
Keywords: | Sub-Project: DNS
Defect Severity: N/A | Estimated Difficulty: 3
Feature Depending on Ticket: | Total Hours: 0
Add Hours to Ticket: 0 |
Internal?: 0 |
-------------------------------------+-------------------------------------
Changes (by jelte):
* owner: jelte => jinmei
Comment:
Replying to [comment:7 jinmei]:
> '''general'''
>
> (Just a comment) Code/tests/document for both opcode and rcode seem
> quite resemble and may suggest unification. But considering this is
> probably tentative experiment it's probably okay.
>
> '''auth.spec.pre.in'''
>
> - 'NXRRSET' is missing within the parentheses?
> "item_description": "The number of total responses with rcode 8
()"
>
ack, added
> '''auth_srv.cc'''
>
> - Some cases are not counted, and some cases (where the response is
> not generated) are incorrectly counted. Former example is this:
Ack, added a private method like your proposed wrapper.
> Also, we need to have tests that would fail without these
> corrections.
>
Naturally. I even added a way to test them more thoroughly than they were
(also checks if for all counters are zero), perhaps we can do something
similar for opcode, but we should probably wait on that possible refactor.
BTW, not all paths are covered; I don't think there is currently a way to
trigger the Exception catchall (the one that causes SERVFAIL after
fromWire on auth_srv.cc:431). I tried to add a wiredata-mangling feature
to our test system but then i discovered that all the actual parsing code
at this point can only throw ProtocolErrors. Oh well.
> '''statistics_unittest.cc'''
>
> There are a couple of trivial typo due to (seeming) copy-paste. I've
> fixed them directly.
>
Thanks
> '''others'''
>
> I think we need a changelog entry for this.
Oh right, of course;
{{{
XXX. [func] jelte
b10-auth now also experimentally supports statistics counters of the rcode
reponses it sends. The counters can be shown as rcode.<code name>, where
code name is the lowercase textual representation of the rcode (e.g.
"noerror", "formerr", etc.).
Same note applies as for opcodes, see changelog entry 364.
}}}
Speaking of which, we should probably also have a ticket that allows you
to do 'Stats show Auth opcode' (i.e. only show opcode, but all of them),
and perhaps even a more general 'don't show zeroes'. Also looking forward
to discussing more general ways to do statistics now, btw :)
--
Ticket URL: <http://bind10.isc.org/ticket/1613#comment:9>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development
More information about the bind10-tickets
mailing list