From: Kapil A. <ka...@cc...> - 2013-06-12 22:23:00
|
Hi Alan, Could you please checkout the latest source from svn: svn checkout svn://svn.code.sf.net/p/dmtcp/code/branches/1.2 dmtcp-1.2 The dmtcp_resart_script.sh now accepts a --ckptdir flag where you can specify the new checkpoint directory. I believe in your case, you would want to use the same directory for both --resartdir and --ckptdir flags to point to the same directory. Thanks, Kapil On Wed, Jun 12, 2013 at 5:17 PM, Gene Cooperman <ge...@cc...> wrote: > Hi Alan, > Thanks for making us aware of this problem. We weren't aware of it. > > We're committed to providing you with a prompt solution to this (in the > next few days). If you don't mind, we'd like to simply send you a tarball, > so that you can verify within your own working environment that this > fixes the problem. > > DMTCP 1.2.7 was released in March of this year. I'm assuming that > the older DMTCP 1.2.6 used to work with Condor. We are planning to shortly > issue DMTCP 1.2.8, and we can add any needed fixes into DMTCP 1.2.8. > The tarball we send to you will be based on the upcoming DMTCP 1.2.8. > > Kapil, > You had originally written the DMTCP_RESTART_DIR capability -- > in part to support Condor. Could you look at any changes that may > have affected it between 1.2.6 and 1.2.7? > > On Tue, Jun 11, 2013 at 02:42:17PM -0500, Alan De Smet wrote: > > We're moving our DMTCP generated checkpoints around before > > restarting them. We use DMTCP_RESTART_DIR to point things at the > > new directory. DMTCP 1.2.7 doesn't seem to like this; it really, > > really wants to write the ckpt_BINNAME_*.ckpt file back in the > > original directory, going so far as to try to recreate the > > containing directory if it's missing. If it fails to write that > > file in the original, DMTCP exits with 99 when it next tries to > > checkpoint. > > > > This kills our ability to use DMTCP under HTCondor. Is this a > > known issue? Is there a fix or workaround? > > > > -- > > Alan De Smet Center for High Throughput Computing > > ad...@cs... http://chtc.cs.wisc.edu > > > > > ------------------------------------------------------------------------------ > > This SF.net email is sponsored by Windows: > > > > Build for Windows Store. > > > > http://p.sf.net/sfu/windows-dev2dev > > _______________________________________________ > > Dmtcp-forum mailing list > > Dmt...@li... > > https://lists.sourceforge.net/lists/listinfo/dmtcp-forum > |