Archive validation on the VO
The inclusion of active resources on the VO website is a relatively new feature and actively being improved. Please use the feedback tab on the right of the website for any feedback or issues you experience.
You can launch an experiment using the 'Design' tab in your group menu, and you can view all experiments in the group using the 'Group Sims' tab. Clicking on the Sim ID number will take you to a more specific status of the experiment.
For the time being, some of the caveats of it's use include:
- The only errors reported are connection errors. To look for experimental failures, you'll need to check the output of your simulation. If you requested output and none was given, either the simulation suffered an unreported failure, or the experiment setup exceeded the run time limitation given. Better error reporting is planned, but is well down the road.
- Max Run Time is a bit of a misnomer. Minimum run time before automatic termination would be a better definition. Experiment status is checked during cron runs, which are every 10 minutes. Cron takes a few minutes to complete, depending on it's tasks. If the current time exceeds the maximum run time setting on the job, cron terminates the experiment automatically. Depending when an experiment is started, it could be {maximum run time} + {cron cycle time} +/- {cron run time variation} before you get results. So, for example, if you set the maximum run time to 2 minutes, it could be 14 or 15 minutes before you get results. So be patient.
- Experiments AND outputs are public to all members of the group. If you want your results to be private to you only, the closest you can come is to closely monitor your experiment for termination, download your output(s), and then delete the experiment. However, control (stop/delete) of the experiment is restricted to the experiment owner, and group/site administrators. The public checkbox when setting up an experiment refers to making access to the outputs available to unauthenticated users.
- Queued jobs are not yet supported. A server must currently be available to run an experiment. We plan to add queuing functionality in the near future. If you get an error that all resources are in use, please try again in a few minutes, or contact us to request additional servers be provisioned if you are having trouble with availability.
- Security cannot be guaranteed. Of course, this is true for any resource accessible to people, but in particular, we are compiling and executing user-supplied source code on our resources. This has some obvious inherent risks, and is why we:
1) Do not run simulations directly on the web server
2) Do not accept commands from the simulation servers on the web server (i.e. the simulation server advertising it's output is ready)
3) Reset the simulation stack for each experiment run
4) Validate submissions manually (not depend on shared simulation servers)
5) ...etc.
This list of caveats will be updated as needed based on questions we receive.