Skip to content

Latest commit

 

History

History
246 lines (203 loc) · 8.29 KB

File metadata and controls

246 lines (203 loc) · 8.29 KB

In Progress

Break

geco_gwpy_dump improvements

Add segment lists to output file list

Segment lists are used to figure out which time segments were good for the detector. These should be considered output files to prevent redundant downloading when unpacking an archived job. Since they are necessary for any work involving a jobspec file anyway, you want to know whether they are there, and it is thus important to list their status when the -o flag is given. It is also therefore important to include them in the archive, so that later data analysis can be performed offline.

Fix incorrect Cesium plot

Used the wrong channel: H1:SYS-TIMING_C_MA_A_PORT_2_SLAVE_CFC_TIMEDIFF_2 is incorrect for LHO Cesium; use H1:SYS-TIMING_C_MA_A_PORT_2_SLAVE_CFC_TIMEDIFF_1 instead.

...
you are analyzing H1:SYS-TIMING_C_MA_A_PORT_2_SLAVE_CFC_TIMEDIFF_2, which
is NTP comparator channel and not the atomic clock. That's why your offset
as well as linear drift is very small. Real atomic clock comparison should
look like the LLO data.

The correct channel to analyze is
H1:SYS-TIMING_C_MA_A_PORT_2_SLAVE_CFC_TIMEDIFF_1.
...
A minor point related to the above is that the diagnostic channel plots like
https://dcc.ligo.org/DocDB/0145/T1700441/001/G297595-
diagnostic-channel-plot-comparison.pdf
should include corner station comparator A channels like
L1:SYS-TIMING_C_PPS_A_SIGNAL_1_DIFF both for LLO and LHO. Corner station
comparator B channels are not connected to anything.
...

Plan

Update jobspec

Change in channels

Change in channel descriptions

Change in plots

Reset plot properties to blank

Update git repo

Start download on ligodl

Start downloading the missing channel on my data download server.

Finished, less than 24 hours for a single channel.

Add utils for easily moving output files around

Plan

Add function for getting sha256 sum of out filenames

Explanation here.

Add function for getting archive filename

Add function for archiving

Explanation here.

Add function for unarchiving

Explanation here.

Add flags

Add -a option for archiving
Add -u option for unarchiving
Add -f option to print archive filename

Update help string with new flags

Add a copy of the jobspec file to the archive

Add a copy of the jobspec JSON file to every archive

Add a function that unarchives an entire job, including spec

Add -x flag to fully extract a jobspec and its output files

Add -X flag, same as -x but without filename checks

Add documentation on -x flag to help string

Add documentation on -X flag to help string

Mention in docs that multiple flags are not officially supported

Resolve symlinks before archiving

Do this using os.path.realpath for all cases.

Resolve symlinks for output files

Resolve symlinks for jobspec file

Copy output files to ldas-pcdev2

Also back up files to research/archive/slow-channel-dumps.

Generate plots

This took way too long and made way too many zoom plots. It also forced the downloading of thousands of files and more than doubled storage used. This is because the LHO Cesium clock was outside of the default anomaly window for most of the run. Just about every anomaly was a plain old outlier.

Adding a new TODO item to fix this now.

Add a CLI to geco_slow_channel_plot

To be honest, a good amount of this was already done in another branch.

Diagnose faults

Update Cesium document

Describe faults

Fortunately there were no faults. Only a spurious missing time at hanford segment 218, as with the other channels.

Update metrics on LHO performance

Update conclusion

No longer need to say that LHO’s superior performance is worth investigating because apparently it isn’t superior after all. Somewhat better calibrated, but not qualitatively so. Similar order of magnitude drift.

Update plots in Cesium document

Send updated file to Zsuzsa

Re-upload to DCC

Waiting on Zsuzsa’s okay.

Wait on Zsuzsa’s clearance

Update geco_slow_channel_plot documentation

Add a line about double checking with people on site.

First of all, put in first line that get in touch with Keith and Dave at
the sites to make sure that no cables were switched...

Done