Question: Why am I seeing "incorrect password" for a machine?
Answer: If the password on your account has changed since you registered the machine such as mpp1, you will need to reconfigure that machine to update your password.
Question: Why does my calc never get past the submitted or running state?
Answer: Network failures or system problems can cause a calculation to stay in the submitted or running state even though the calculation actually runs successfully. In this case the communication of the status/results of the running calculation will not be received by ECCE and therefore updates are no longer performed by the ECCE interface. The calculation results file will exist on the launch machine and can be imported into ECCE once the calculation has completed.
Question: Why did my calculation fail?
Answer: There is no simple answer to explain why a calcultion failed. It could be the result of any number of problems including code installation problems, a syntactically or semantically incorrect input file, or the calcution may simply have failed to converge. This section outlines a process for discovering why a calculation failed.
- Look at the ouput file
- The most common failure is the failure of a calculation to converge. Note that this condition does not necessarily result in the unconverged state. The unconverged state applies to the outer most convergence loop. For example, in a geometry optimization, if the geometry optimization fails to converge, the state will be "unconverged." However, if the initial SCF convergence fails, the state will be "failed."
- Although we attempt to test all conditions, there could be an error in the input file generation that is detected as a syntax error by the code.
- The Calcution Editor is designed to minimize the chances that a syntactically correct but semantically incorrect input file can be generated. However, it is still possible for this to happen for a number of reasons:
- The user used final edit to modify the generated input file.
- There is an error in the way a code is registered within ECCE.
- Insufficient information is registered within ECCE. An example of this is that regarding basis sets, ECCE knows whether the code supports spherical polarization components. However ECCE does not know that NWChem only supports spherical for certain theories. In this situtaion the ECCE tools will allow you to specify options that are not really supported. If you run across a situation like this, please report it using the NWChem wiki ECCE forums.
- To check for these conditions, examine the output file. In the Calculation Manager and Calculation Viewer, there is a "View Output File" menu item. If this option is not enabled, there is no output file so the calculation never ran - proceed to step #2. Otherwise, go to the bottom of the output file. Some information about the failure will be at or near the end of the file. You will need to scroll up to determine the exact cause. Note that any failure with Gaussian94 results in a core dump. You still need to page up to find out what caused Gaussian94 to abort. If you discover that the problem is due to a bug in ECCE, please report it using the NWChem wiki ECCE forums. Otherwise, modify the calculation as necessary and run it again.
In the Calculation Manager and Calculation Viewer, there is a "Run Mgmt" menu th at has a menu item to open a shell in the run directory. If you select this opt ion, you will be able to examine all the files associated with the calculation.
If all else fails, try to run the calculation completely outside of ECCE. Possible problems include:
- The wrong binary for the platform was provided.
- Shared libraries are not accounted for in the LD_LIBRARY_PATH.
- There are permission problems on the executable or related files.
Question: I expected to see a number of processors statement when I performed a final edit on the input file within the Calculation Editor. Why is this statement missing?
Answer: The number of processors set in the Job Launcher will not appear in a "Final Edit" of the input file in the Calculation Editor. This field is added into the chemistry code input file during the launch process and not when the Calculation Editor generates the input file for Final Edit. It isn't necessary to manually set this parameter during Final Edit.
Question: Why does the Job launcher list queue names that are unauthorized?
Answer: In the Job Launcher the list of queue names for a machine denotes all the queues for that machine--not just the ones that the user is authorized to use. It is possible to select a queue that the user is not authorized for and the job will fail when launched for this reason. It is the user's responsibility to know which queues they are authorized to use.
Question: How can I restart a NWChem run with the same ECCE calculation?
Answer: Use the Edit menu item "Reset for Restart" that is available in the Calculation Manager.
Question: Can I setup a new calculation using an intermediate geometry step from another calculation?
Answer: Yes. To accomplish this, you can select the desired geometry step from the geometry trace interface in the Calculation Viewer by clicking on the step in the graph or advancing the step with the arrow buttons. Then use the drag icon in the same interface to drag the selected geometry (using the middle mouse button) into the Builder. As a shortcut there is a "Duplicate with last Geometry" menu item in the Edit menu of the Calculation Manager which creates a new calculation with the final geometry of the selected calculation.
Question: Can I import the results of a run restarted from a NWChem checkpoint file?
Answer: It is possible, but you need the initial NWChem ECCE output file to be prepended to the result file (ecce.out) produced by the checkpoint run. This guarantees that ECCE will find the chemical system and basis set information in a parsable form as well as finding the resultant properties.
Question: Why can't I import the results of my NWChem run?
Answer: To successfully import the results of a NWChem calculation into ECCE, the NWChem calculation must have been setup to create an ECCE parsable output file in addition to the standard NWChem output file. This is controlled by a flag in the NWChem input file. When a NWChem calculation is setup and executed within the ECCE environment this flag is always set to true and parsable output will always be generated. ECCE's default name for this file is "ecce.out." If a user runs NWChem outside of ECCE then the user must manually set this flag to true. The name of the parsable file is controlled by the user. The default NWChem output file (nwchem.out) is not parsable by ECCE and cannot be imported into ECCE.