VE-320: Fix dual backend balancing in pulsar modes (COMPLETE)
Offline script validation (INCOMPLETE)
Verify that LBW gain balancing works in Matrix spectral line modes (COMPLETE)
Implement messaging/reporting of LBW gain balancing errors (COMPLETE)
Enable automatic switching between classic and matrix HPC modes (IN PROGRESS)
Investigate packet loss in c1500 modes (INCOMPLETE)
High time resolution modes (INCOMPLETE)
Final spectral line regression tests (INCOMPLETE)
Final pulsar mode tests (INCOMPLETE)
Continue work on pulsar Matrix
Append delays to BOF file names
Story Details
VE-320: Fix dual backend balancing in pulsar modes
Ray added a conditional in the balancing routine that skips any attempt at LBW gain balancing in the pulsar modes.
A Balance() command when configuring for both GUPPI and Astrid only balances the GUPPI IF path, but an additional Balance("VEGAS") balances the VEGAS IF and no longer causes any error messages or abort attempts in Astrid. Ray considers this a perfectly acceptable solution.
Offline script validation
No updates
Richard reported this, but Ryan thinks that the exact issue is that validation fails in offline mode when on the regression candidate branch for spectral line modes.
This may also be an issue for pulsar modes, but Ryan has not rigorously verified this.
Verify that LBW gain balancing works in Matrix spectral line modes
Ryan believes that Richard checkes this on Thursday on banks for which the IF balancing was OK.
Ray has verified that this works in the regression environment.
Implement messaging/reporting of LBW gain balancing errors
During testing on Thursday Richard, Ray, and Ryan confirmed that a CLEO error is now issued if VEGAS fails to balance. However, there were obvious issues with the IF balancing that seem to be related only to the regression test candidate branch. These will need to be investigated. Another round of testing to make sure the CLEO messages appear in the correct context may be needed.
If the balancing of any component reaches the maximum number of attempts a message is issued.
Enable automatic switching between classic and matrix HPC modes
Ray has just started work on this, and is in the debugging stage.
Investigate packet loss in c1500 modes
Ryan will prioritize talk with Jason and Randy about this.
Continue work on pulsar Matrix
Joe got pulled away on this to work on the VEGAS spectral line day display.
Fix VEGAS spectral line data display
The web-based display uses an interface to the manager to stream decimated data. In classic mode the manager wrote the FITS files itself and had the necessary information.
In the Matrix the HPC writes the FITS files instead of the manager.
Joe is implementing a Matrix interface to enable this. This is needed for any Matrix spectral line release.
Append delays to BOF file names
No updates.
High time resolution modes
No updates.
Final spectral line regression tests
Waiting on fixes to IF balancing, matrix/classic switching.
Final pulsar mode tests
Waiting for freeze of pulsar modes.
VEGAS Pulsar Project Meeting - Sprint Planning: 2017 August 14th 2:00 - 3:00pm ET
Note: The team's "velocity" has been consistent at 11--13 points per sprint. We should aim for 12 points per sprint for planning purposes.
Backlog
High time resolution modes
Investigate packet loss in c1500 modes
Continue work on Matrix LBW pulsar modes
Critical Issues
Offline script validation
Enable automatic switching between Matrix and classic modes.
Continue working on Matrix spectral line data display
New Stories
MAC addresses for 10-gig ethernet ports are wrong for Banks C and D in VPM coherent modes
The MAC addresses seem totally invalid (10.10.10.11)
Ray needs to see if he can recreate this in the regression environment
Ray suggests setting up logging on the ROACH to see when this happens.
Identify a source tree for VPM utilities
Install VPM utilities in new source tree
Add a Task Master task for vpm_auto_plot, automatic VPM bandpass monitor, and a DAQ server used to monitor processes on the VPM HPCs in coherent dedispersion modes.
Investigate IF balancing in Matrix spectral line modes
Banks E, F, G, and H failed to properly balance at the IF rack during testing on Thursday. This was persistent across different spectral line modes. We didn't have a chance to investigate in detail, but the issue did not appear in a quick check of a handful of modes when we moved back to 16.3, so it would appear to be related to the regression candidate specifically.
The exact symptom was too little power in at least one polarization channel on the above banks. The IF was set up for X-Band.
In addition, one of the spectral windows in Bank A was cut off at the low end of the band. It almost looked as if a filter was not centered properly, but this is pure speculation. Again, this did not appear after switching back to 16.3