BIND 10 #299: AXFR fails half the time

[BIND 10 Development]

#299: AXFR fails half the time
-----------------------------+----------------------------------------------
      Reporter:  zzchen_pku  |        Owner:  zzchen_pku
          Type:  defect      |       Status:  reviewing 
      Priority:  major       |    Milestone:            
     Component:  xfrout      |   Resolution:            
      Keywords:              |    Sensitive:  0         
Estimatedhours:  0.0         |        Hours:  0         
      Billable:  1           |   Totalhours:  1.5       
      Internal:  0           |  
-----------------------------+----------------------------------------------

Comment(by zhanglikun):

 Replying to [comment:30 jinmei]:
 > I'm not sure if we are talking about the same thing.  What do you mean
 by "tarnsfer-in"?  I've been exclusively talking about "transfer-out",
 where a secondary server sends an AXFR request to a BIND 10 primary server
 and BIND10 replies to it via the xfrout process.
 >

 Sorry, my mistake, I wrote wrong word, it should be 'transfer-out', we
 were and are talking about the same thing.

 > Now, consider BIND 10 manages two large zones and the following sequence
 of events:
 >
 > 0sec: client A sends an AXFR request for zone X.
 > 0.1sec: b10-xfrout starts handling the request.  it takes 10 minutes.
 > 0.2sec: client B sends an AXFR request for zone Y.  It won't get any
 response soon because b10-xfrout is busy for client A.
 > 10min 0.1sec: b10-xfrout complete the transfer for client A, and starts
 handling client B. (but client B may have given up the transfer by then)
 >

 the events should like this:

 =0sec: client A sends an AXFR request for zone X.
 =0.1sec: b10-xfrout starts handling the request by creating one thread,
 then b10-xfrout starts to listen to the next AXFR request. it takes 10
 minutes for zone X's transfer-out, then the created transfer-out thread
 terminate.
 =0.2sec: client B sends an AXFR request for zone Y, b10-xfrout create
 another thread for zone Y's transfer-out. Now there are two transfer-out
 threads.

 =0.2+sec: transfer-out for client B finish, transfer-out thread
 terminates. There is only one transfer-out thread now(for client A).

 =10min 0.1sec: transfer-out for client A finish, no transfer-out thread
 now.

 > Are we on the same page?

 Yes, we are.

-- 
Ticket URL: <http://bind10.isc.org/ticket/299#comment:31>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development