Fail to reserve resources on a Grid5000 site is the most frequent failure in launching the Demo. If the user see the error message like this:
Reservation not valid --> KO (This reservation may be run at 2013-10-18 07:01:01) No resource available on G5K, please try the demo later.
he can either install the Demo on another site and launch it again, or wait some time until resources are available.
Sometimes even the reservation is accept by Grid5000, the scheduled start time may be postponed according to the state of the site. If the reservation is postponed, the Demo may fail to find available nodes, and show the error message as below:
***** Retrieve available nodes. ***** XXXXXX Job id is XXXXXX Temporary file to store available nodes is created. **ERROR: Not nodes available.
In this case, the user can wait for a while and use the
-d options to re-launch the Demo.
~/BlobSeer-Demo/demo-scripts/demo-launch.sh -i <job-id> -d true
If a chained test, or a single test is failed, it is possible that Hadoop-BSFS/HDFS is still deployed on Grid5000 nodes. The previous deployment may prevent the execution of the new test. In this case, the user can run the script
cleanup-test.sh to cancel the deployed Hadoop-BSFS/HDFS.
If the user breaks something on the Grid5000 nodes so that Hadoop-BSFS/HDFS cannot be launched correctly, the user can leverage the
-d option of
demo-launch.sh to recover the nodes to the initial state.
Some times the errors occur during a chained test may cause the failure or abnormal behavior of the visualization function. Usually, the visualization function will stop after a certain period depends on the size of the chained test. However, if the user is not patient to waiting till the end, he can use
Ctrl-C to quit the program directly. Then, the script
cleanup-inc.sh is used to kill the gnuplot drawing process.