Issues following 9.3.5-p1 upgrade to 9.7.0-p1 Windows VMware environment

[Danny Mayer]

rsmith at osc.gov.on.ca wrote:
> We have been running Bind 9.3.5-P1 on Windows 2003 guest in VMware ESX
> 3.5 environment for many years with no issues. Following an upgrade to
> Bind 9.7.0-p1 we are experiencing a couple of issues. No Bind
> configuration changes were made to config files other than the
> allow-query-cache, additional-from-auth and additional-from-cache
> statements so that queries worked in the new version. The errors are
> occurring on multiple Bind servers on different Vmware ESX hosts.
> 
> On a regular but random basis we are getting the 2 socket error messages
> logged. I have not been able to determine what is causing this to occur
> nor reproduce at will.
> 
> .\socket.c:2444: unexpected error:
> SOCKET_RECV: Windows error code: 1236, returning ISC error 54
> 
> We are also getting a poked timer error logged consistently soon after
> Bind service start and infrequently after that.
> 
> *** POKED TIMER ***
> 
> Any suggestions / solutions would be greatly appreciated.

This looks like a bug in the code. ISC error 54 means that it got a
connection reset on the receive for the packet and Windows error code
indicates that "The network connection was aborted by the local system"
which means that BIND canceled the receive and this is the tail end of
handling the cancellation. Unlike Unix, Windows handles all of this
asynchronously, so it's not completed until it reaches this set of code.

This looks and smells like a bug to me. Please file a bug report by
sending email to bind9-bugs at isc.org with these details.

Danny