[QPID-3686] Donor/Donee Brokers should not accept connections?

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Won't Fix
• Affects Version/s: 0.10
• Component/s: C++ Broker
• Labels: connection session timeout

Environment

cat /etc/redhat-release
CentOS Linux release 6.0 (Final)

uname -a
Linux ip-10-196-231-143 2.6.32-71.29.1.el6.x86_64 #1 SMP Mon Jun 27 19:49:27 BST 2011 x86_64 x86_64 x86_64 GNU/Linux

rpm -qa qpid-cpp-server*
qpid-cpp-server-0.10-8.el6_1.x86_64

• Assignee: Unassigned
• Reporter: Joel Caplin
• Created: 14/Dec/11 21:38
• Updated: 15/Dec/11 19:28
• Resolved: 15/Dec/11 19:27

Description

TO REPRODUCE:

Suppose I have three C++ brokers (q1, q2, q3) in a cman/corosync-managed cluster. I have a java client J1. Then:

t0 - all three brokers are up
t1 - q1 falls over
t2 - q1 is restarted; catch up begins with q2 select as donor
t3 [i] - EITHER J1 starts; it connects to q1 and attempts to create a session
t3 [ii] - OR J1 starts; it connects to q2 and attempts to create a session
t3 + 60000 milliseconds - J1 raises SessionException (see attached stack trace)

It appears that the connection is effectively blocked from doing anything useful while the donor/donee is participating in a catch up.

Would it be possible to reject connections until such time that the broker is able to fulfill requests (i.e. able to create a JMS sesssion from my client's perspective?)