[Condor-users] upcoming version 6.8.x releases

[Todd Tannenbaum <tannenba@xxxxxxxxxxx>]

I just answered this question directly for a colleague; I figured the 
information could be useful to folks on this list as well.  Of 
course, all this info is subject to change, but it reflects our current plan.

regards,
Todd


> Todd,
>
> Any estimates of when 6.8.1 may come out? Would it be a matter of
> weeks or months? I'm considering whether it is worth to migrate
> from 6.6.x to 6.8.0 or simply wait for 6.8.1 if that's coming out
> soon.
>
> Tony

Hi Tony --

It is a matter of weeks; hopefully as soon as end of next week.  We 
already have v6.8.1 code complete, but it is still being 
tested.   Below is a list of the major bug fixes so you can decide 
for yourself if you wish to wait for v6.8.1.   Of course, since it is 
a stable series, there are only bug fixes - no significant new 
features (since that could introduce new bugs!).

Version 6.8.2 will come along about a week or two after v6.8.1.  The 
only change in v6.8.2 will be support for checkpoint/restart 
(standard universe) on x86_64 chips running Linux.  (i.e. 
checkpointing of 64-bit processes on 64-bit Linux distros).

Hope this info is helpful,
regards,
Todd

* Fix issues w/ checking to see if a user log file is on NFS. 
Reported by NMI running Condor's regression test suite.
* Fix Kerberos scalability problems.  Reported by CDF.
* Fix CEDAR OpenSSL authentication.  Reported in condor-users.
* Fix bug on Win32 where removal of a DAGMan job does not remove job 
submitted by that dag (works fine on Unix).  Reported in condor-users.
* Fix bugs related to non-blocking connects that can cause daemons to 
EXCEPT after running out of file descriptors.  Reported by GLOW.
* Fix bugs related to non-blocking connects that can cause daemons to 
crash.  Reported by Fermi.
* Fix race condition bugs with disconnected starter/shadow (i.e. job 
leases).  Discovered by Todd.
* Fix bugs with condor_rm -f, and add administrator control over its 
use.  Reported by LIGO.
* With grid universe, if remote job has a hard lease, don't bother 
putting a job on hold if errors encountered during job 
removal.  Reported by gLite.
* Investigate and/or fix issues with the master failing to kill 
DAGMan processes if the schedd crashes.  Reported by LIGO.



-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Todd Tannenbaum                       University of Wisconsin-Madison
Condor Project Research               Department of Computer Sciences
tannenba@xxxxxxxxxxx                  1210 W. Dayton St. Rm #4257
http://www.cs.wisc.edu/~tannenba      Madison, WI 53706-1685
Phone: (608) 263-7132  FAX: (608) 262-9777