Hello, Thanks for the reply, I updated the code repository and was trying to get it all working again. (Had to change my DMTCP event to the USER_THREAD_RESUME apparently) I noticed a bug within dmtcp/jalib/jfilesystem.cpp, the dirname function does not work with files placed in the root "/xxxx". A check if the lastSlash is zero and if so returning a slash would solve this issue.
Then my own problem. This required a lot of testing :( But I have now figured out that whenever I checkpoint a VM, it does not save the /tmp folder. This is a problem since DMTCP saves it's information in that folder. I noticed as well that dmtcp_restart_script.sh did not have the flag to change the temp folder. So I thought it was not possible. It's only recently that I discoverd dmtcp_restart to have such a flag ! When I changed the DMTCP_TMP directory it was fixed. I now can take as many DMTCP checkpoints and restarts as I want without problems. Therefor I would like to request if this could be added to the restart script. As an attachment I have added a patch that can be used as well. I hope it helps, thanks again for all your support :) Robin 2013/3/29 Kapil Arya <[email protected]> > Hi Robin, > > I tried to reproduce your bug but couldn't do it. Are you using DMTCP > trunk? Also, are you using the new svn url: > svn co svn://svn.code.sf.net/p/dmtcp/code/trunk dmtcp-trunk > > The old one is now obsolete. > > In the past few days, I have pushed a few changes into the trunk. Can > you checkout the latest version and let me know if the problem still > persists. If it does, we need to come up with a way to reproduce the > failure to debug. > > Kapil > > On Sun, Mar 24, 2013 at 4:19 PM, robin staes <[email protected]> > wrote: > > Hi, > > > > I have been working with DMTCP quite a lot recently and have encountered > > several issues which were resolved by your excellent support. > > Having encountered another problem I hope to benefit from this support > once > > more. > > > > Currently I am working on a system that should robustly checkpoint a > process > > that runs on a virtual machine. > > This is done by checkpointing the application and through the usage of > the > > plugin system with which I checkpoint the VM itself. (This is via a > "system" > > call, which I appreciate you have implemented!) > > This is currently working but not completely. > > My process: > > > > Start a simple python counter and let it checkpoint every 10 minutes. > > After a checkpoint kill the VM and reboot it from the VM checkpoint. > > This works perfectly. > > But when the restarted process takes a new checkpoint and I want to > restart > > it, it fails. > > The error code 99 is returned which indicates a problem within DMTCP. > > After having recompiled to enable debug information I have traced the > > problem to this area: > > > > [40000] TRACE at pid.cpp:92 in openOriginalToCurrentMappingFiles; > > REASON='Open dmtcpPidMapFile' > > pidMapFile.str() = > > /tmp/dmtcp-root@ip-10-194-33-122 > /dmtcpPidMap.13a71f78f7523f34-40000-514f32c7.514f57098 > > [40000] ERROR at pid.cpp:80 in openSharedFile; REASON='JASSERT(false) > > failed' > > name = > > /tmp/dmtcp-root@ip-10-194-33-122 > /dmtcpPidMap.13a71f78f7523f34-40000-514f32c7.514f57098 > > strerror((*__errno_location ())) = No such file or directory > > Message: Cannot open file > > python2.7 (40000): Terminating... > > > > It is true that when I look at the designated folder I can't find the > file. > > But my first restart works and the pid.cpp:92 result is: > > > > [40000] TRACE at pid.cpp:92 in openOriginalToCurrentMappingFiles; > > REASON='Open dmtcpPidMapFile' > > pidMapFile.str() = > > /tmp/dmtcp-root@ip-10-194-33-122 > /dmtcpPidMap.13a71f78f7523f34-40000-514f32c7.514f56bb5 > > [40000] TRACE at virtualidtable.h:241 in writeMapsToFile; REASON='Write > Maps > > to file' > > mapFile = > > /tmp/dmtcp-root@ip-10-194-33-122 > /dmtcpPidMap.13a71f78f7523f34-40000-514f32c7.514f56bb5 > > > > Now as well I can't find the file on my filesystem. > > > > I hope you can enlighten me. > > > > Thanks in advance, > > > > Robin Staes > > > > PS: Sorry for the mail bombardment to Kapil, I did not have the nerve to > > send another one :) > > > > > > > ------------------------------------------------------------------------------ > > Everyone hates slow websites. So do we. > > Make your web apps faster with AppDynamics > > Download AppDynamics Lite for free today: > > http://p.sf.net/sfu/appdyn_d2d_mar > > _______________________________________________ > > Dmtcp-forum mailing list > > [email protected] > > https://lists.sourceforge.net/lists/listinfo/dmtcp-forum > > > -- Robin
dirname_and_restartscript_fix.diff
Description: Binary data
------------------------------------------------------------------------------ Own the Future-Intel(R) Level Up Game Demo Contest 2013 Rise to greatness in Intel's independent game demo contest. Compete for recognition, cash, and the chance to get your game on Steam. $5K grand prize plus 10 genre and skill prizes. Submit your demo by 6/6/13. http://altfarm.mediaplex.com/ad/ck/12124-176961-30367-2
_______________________________________________ Dmtcp-forum mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/dmtcp-forum
