BIND 10 #1271: exception on extremely slow startup

[BIND 10 Development]

#1271: exception on extremely slow startup
--------------------------------------------+-----------------------------
                   Reporter:  jreed         |                 Owner:
                       Type:  defect        |                Status:  new
                   Priority:  minor         |             Milestone:
                  Component:  Unclassified  |            Resolution:
                   Keywords:                |             Sensitive:  0
            Defect Severity:  N/A           |           Sub-Project:  Core
Feature Depending on Ticket:                |  Estimated Difficulty:  0
        Add Hours to Ticket:  0             |           Total Hours:  0
                  Internal?:  0             |
--------------------------------------------+-----------------------------

Comment (by jinmei):

 Replying to [comment:1 shane]:
 > Even though the report is that it took over two minutes, the program
 thought it only took 9 seconds.
 >
 > I'm not sure what we can do on a machine so badly broken. :(

 This exact same happened on my personal server.  (You might call it
 broken:-) it's (in today's standard) unusually slow, but otherwise
 it works perfectly fine.  In my case sleeping for 1 second wasn't
 sufficient for cfgmr to start up:

 {{{
     def start_cfgmgr(self, c_channel_env):
         """
             Starts the configuration manager process
         """
 [...]
         # sleep until b10-cfgmgr is fully up and running, this is a good
 place
         # to have a (short) timeout on synchronized groupsend/receive
         # TODO: replace the sleep by a listen for ConfigManager started
         # message
         time.sleep(1)
 }}}

 and so it encounters "broken pipe" in start_ccssession:

 {{{
         self.start_cfgmgr(c_channel_env)
         self.start_ccsession(c_channel_env)
 }}}

 I tentatively changed the sleep time to 5 seconds, and it worked,
 but it's needless to say that it's just an ugly workaround.

 As noted in TODO, what the boss process should actually do is to establish
 more explicit synchronization with cfgmgr.  It should wait on msgq,
 cfgmgr sends an "I'm ready" message to the boss when it starts up,
 the boss continues its process then.  We might also want to specify
 a limit on the waiting time (and since it's before the boss gets the
 configuration, it must be given via a command line option or an
 environment variable).

 I'd also suggest revising the log messages in this failure case.  It's
 difficult to know what was actually wrong from these messages.  Seeing
 the bare 'broken pipe' exception is also not cool.

 Finally, not directly related to this issue, but this log message is
 not good:
 {{{
 2011-09-29 15:39:49.376 INFO  [b10-boss.boss] BIND10_STARTING_PROCESS
 starting process ccsession
 }}}
 because ccsession is not a "process".  This is a result of a naive
 copy-paste here:
 {{{
         self.log_starting("ccsession")
 }}}
 it should be fixed.

 I proposed this ticket to be handled in the next sprint.

-- 
Ticket URL: <http://bind10.isc.org/ticket/1271#comment:3>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development