BIND 10 #2137: update the boss module according to the new statistics model

[BIND 10 Development]

#2137: update the boss module according to the new statistics model
-------------------------------------+-------------------------------------
                   Reporter:         |                 Owner:  naokikambe
  naokikambe                         |                Status:  reviewing
                       Type:         |             Milestone:
  enhancement                        |  Sprint-20120821
                   Priority:         |            Resolution:
  medium                             |             Sensitive:  0
                  Component:         |           Sub-Project:  DNS
  statistics                         |  Estimated Difficulty:  4
                   Keywords:         |           Total Hours:  0
            Defect Severity:  N/A    |
Feature Depending on Ticket:         |
        Add Hours to Ticket:  0      |
                  Internal?:  0      |
-------------------------------------+-------------------------------------
Changes (by jelte):

 * owner:  jelte => naokikambe


Comment:

 > Replying to [comment:7 jelte]:
 > > The changes in get_stats_data() in boss make the stats daemon quit
 during startup. That obviously needs to be fixed :)
 > Sorry, I couldn't understand that. Do you mean the new _get_stats_data()
 method can stop the stats daemon starting? Could you explain how should I
 fix? In the former code, boss might have accidentally stopped because of
 the wrong format.
 >

 The problem is that the new _get_stats_data() makes a response message of
 the form
 { "boot_time": <some time> }

 And the stats daemon expects the format
 { "owner": <owner>,
   "data": <data>
 }

 So currently, upon startup of bind10, the stats daemon immediately
 crashes:

 {{{
 2012-08-13 15:37:04.438 INFO  [b10-stats.stats] STATS_STARTING starting
 Traceback (most recent call last):
   File "/home/jelte/repos/bind10/src/bin/stats/b10-stats", line 503, in
 <module>
     stats.start()
   File "/home/jelte/repos/bind10/src/bin/stats/b10-stats", line 166, in
 start
     args["owner"], **args["data"])
 KeyError: 'owner'
 2012-08-13 15:37:04.446 INFO  [b10-boss.boss] BIND10_PROCESS_ENDED process
 18047 of b10-stats ended with status 256
 }}}

 > > But also, it is probably a good idea not catch formatting errors in
 the stats daemon, and log an error if that happens (but to keep running)
 > Do you mean boss should check the format of statistics data being sent
 out?  I don't think the validation is always needed because the format is
 constant in boss.  And should it log if it is invalid? The stats module
 logs an error when it receives the wrong format from boss.

 Oh, I don't think the boss should necessarily check its output, but I'd
 check in stats.py.in whether the 'args' value does indeed contain "owner"
 and "data" before continuing (and log an error otherwise);
 e.g. something like either
 {{{
     if "owner" in args and "data" in args:
         <process>
     else:
         logger.error(blabla)
 }}}

 or
 {{{
     if "owner" not in args or "data" not in args:
         raise StatsError("Badly formatted response to get_stats, owner or
 data missing")
 }}}
 but in the latter case, I'd make a raised statserror not fatal (but catch
 it, log it as an error, and continue to run)

-- 
Ticket URL: <https://bind10.isc.org/ticket/2137#comment:9>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development