Known Issues for the VLA Pipeline in CASA 5.1.0
About this Page
This page describe issues affecting the CASA Integrated Pipeline in CASA 5.1.0. These issues are known to either cause the pipeline to fail mid-process, causing the scripts to exit to the CASA prompt or the unexpected exiting of CASA itself, or to cause problems for calibration quality. These failures may result in little or no calibration and potentially corruption of the MS in some cases.
Last update: 2018-12-12.
Testdelay.k does not exist
The symptom is the pipeline fails to complete and the last line in the casa log file is:
INFO gaincal::::casa Exiting script: Table testdelay.k does not exist
This may be caused by the pipeline finding no unflagged data on the delay calibrator, causing all subsequent calibration steps to fail. You may still be able to calibrate these data by hand or with the pipeline by selecting a different field as the delay calibrator.
Extra flagging of edge channels
The symptom is the pipeline flags on extra edge channel (lowest frequency) of each spw with each of three calls to applycal in the pipeline due to a bug in way linearflag with frequency interpolation was implemented. This affects the release version of CASA 5.1.0 available on CASA web page as no patch is currently available. The current resolution is a switch from linearflag interpolation over frequency to nearestflag. The fix to linearflag itself is planned to be released along with CASA 5.3.0 in the future. To get the fixed version of CASA 5.1.0 used for the production pipeline and VLASS, please download this version of CASA 5.1.0. This version will have pipeline version 40844 instead of the standard release version 40833.
Segmentation fault during tclean
The symptom is the pipeline fails with a segmentation fault near the very end of the pipeline during one of the calls to tclean. We are currently still investigating this issue, but a workaround involves running the pipeline without calibrator imaging by commenting the call to both hif_makeimlist and hif_makeimages from any compatible casa_pipescript.py file. For information about running the pipeline via a casa_pipescript.py file, please see the main CASA pipeline web page.
Memory leak and eventual hung process in hif_makeimlist
The symptom is the pipeline tries to take all available memory and eventually freezes the pipeline CASA task somewhere during hif_makeimlist. This issue is currently being investigated and appears to be related to a specific sequence of fully flagged spws. A solution is in development stage now, but a workaround involves running the pipeline without calibrator imaging by commenting the call to both hif_makeimlist and hif_makeimages from any compatible casa_pipescript.py file. For information about running the pipeline via a casa_pipescript.py file, please see the main CASA pipeline web page.
Two sources listed with Bandpass and/or Delay Scan Intents
If an observation has two or more sources listed with a scan intent for bandpass calibration (or delay calibration), the pipeline will try to use both sources for bandpass calibration, applying a time-variable bandpass calibration depending on how close in time the phase calibrator scans are to the different bandpass calibrators. The same is true of delay calibrator scan intents. A solution is to edit the Scan.xml file within the SDM-BDF prior to running the pipeline to remove the extra bandpass or delay scan intents or to split the data by which bandpass calibrator is desired.
Data with TARGET intents on a calibrator source
The VLA pipeline in 5.1.0/5.1.1 will not work when a field has an intent as a calibrator (CALIBRATE_AMPLI, CALIBRATE_PHASE, CALIBRATE_FLUX) as well as a target (TARGET), even when the scans for the intents are separated. This situation is frequently the case in older data when the scan for setting the requantizer gains was given a TARGET intent. We anticipate a fix in CASA 5.4. and recommend to either change the scan intents or to calibrate the data with the CASA 4.7.2 VLA pipeline.
Issues not covered here
If you are experiencing issues running the pipeline that cause a failure to complete or if you have questions about the pipeline, please contact us via the Pipeline Department of the NRAO Helpdesk.
