Hi, Everyone! I am testing DMTCP in order to understand how it works. I am using DMTCP version 2.4.1 version I have three questions:
1) application initiate checkpoint Upon executing the file ./test/plugin/applic-initiated-ckpt/applic I get the following message: ---- *** dmtcp_is_enabled: executable seems to not be running under dmtcp_launch. *** dmtcp_checkpoint: DMTCP is not running. Skipping checkpoint. *** Process done executing. Successfully exiting. ----- If I understood correctly this file should execute a checkpoint after a time and it neither need a coordinator nor to be executed with dmtcp_launch. It is right? I have set my DMTCP_CHECKPOINT_INTERVAL=5. I also compile it through the makefile, that is, it is using the -fPIC. 2) I am testing a simple mpi application: 2 processes increments a value each second. I am using openmpi version 1.7 dmtcp_launch mpiexec -np 2 ./test But DMTCP only creates the checkpoint manually, that is, pressing c<return> on the coordinator. I have set my DMTCP_CHECKPOINT_INTERVAL=5, then, if I understood correctly, DMTCP should creates a checkpoint every 5 seconds, it is right? I have also tested the same example with version dmtcp-1.2.8 and it works fine, that is, creates checkpoint every 5 seconds. 3) When I use the option --no-coordinator I have the following error message: [8215] ERROR at util_init.cpp:42 in getCoordHostAndPort; REASON='JASSERT(mode & COORD_JOIN || mode & COORD_NEW || mode & COORD_ANY) failed' dmtcp_launch (8215): Terminating. There is some particularity that I have missed to use the option --no-coordinator? I would like to use DMTCP with mpi, but I will need to use the option --no-coordinator in order to the application makes checkpoints independently, that is, with no coordinator. Thank you in advance! Edson
------------------------------------------------------------------------------
_______________________________________________ Dmtcp-forum mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/dmtcp-forum
