Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: None
-
Labels:None
-
Story Points:4
-
Epic Link:
-
Team:Data Facility
Description
Reprocess the HSC RC1 dataset, as defined in DMTR-31 Sect. 4.1 , using Stack w_2018_03. Steps include makeSkyMap.py, singleFrameDriver.py, mosaic.py, coaddDriver.py, and multiBandDriver.py (DM-10129DM-11020
Summarize the total node-hours needed.
Attachments
Issue Links
Activity
After checking through the coadd log files, I have found that quite a few visits were removed because the median e residual was too large and their scaled size scatter was too large. A list of these visits is included below.
| coadd visit list & filter | number of "Removing visits" found |
|---|---|
| Cosmos-G | 53 |
| Cosmos-I | 86 |
| Cosmos-N | 16 |
| Cosmos-R | 294 |
| Cosmos-Y | 937 |
| Cosmos-Z | 264 |
| Wide-G | 81 |
| Wide-I | 132 |
| Wide-R | 27 |
| Wide-Y | 127 |
| Wide-Z | 33 |
This information was found with the command grep -c 'Removing visit' coadd*.
The number of occurances for Cosmo-Y seem incredibly high, but after looking at the log files in detail, it would seem that if a visit & ccd pair are complained about once and put on the "removing visits" list, then they are complained about more than twice in the log file. I am looking further into why this happened.
Samantha Thrush
added a comment - - edited After checking through the coadd log files, I have found that quite a few visits were removed because the median e residual was too large and their scaled size scatter was too large. A list of these visits is included below.
coadd visit list & filter
number of "Removing visits" found
Cosmos-G
53
Cosmos-I
86
Cosmos-N
16
Cosmos-R
294
Cosmos-Y
937
Cosmos-Z
264
Wide-G
81
Wide-I
132
Wide-R
27
Wide-Y
127
Wide-Z
33
This information was found with the command grep -c 'Removing visit' coadd* .
The number of occurances for Cosmo-Y seem incredibly high, but after looking at the log files in detail, it would seem that if a visit & ccd pair are complained about once and put on the "removing visits" list, then they are complained about more than twice in the log file. I am looking further into why this happened. After running wc -l to find calexp files in the deepCoadd-results directory, I found that the Wide tracts 8767 and 8766 both have 81 patches in all 5 filters, thus aren't missing any patches. However, tract 9813 is missing some patches in each filter:
- HSC-G: 78 patches (missing:{0,0}, {0,8}, {8,0})
- HSC-I: 79 patches (missing:{0,0}, {0,8})
- HSC-R: 74 patches (missing:{0,0},{0,1},{0,7},{0,8},{1,8},{8,0},{8,8})
- HSC-Z: 79 patches (missing:{0,0}, {0,8})
- HSC-Y: 79 patches (missing: {0,0}, {0,8})
- NB0921: 76 patches (missing: {0,0}, {0,7}, {0,8}, {1,8}, {8,8})
This closely mirrors the results shown in DM-11345 for all 9813 filters except for HSC-Z (that run was missing the {8,8} patch in addition to those mentioned above).
After checking to see which patches are included in deepCoadd-results/merged (created by multiBandDriver.py), I found that tracts 8766 and 8767 included all 81 patches, while tract 9813 included 79 patches (missing: {0,0}, {0,8}).
Samantha Thrush
added a comment - After running wc -l to find calexp files in the deepCoadd-results directory, I found that the Wide tracts 8767 and 8766 both have 81 patches in all 5 filters, thus aren't missing any patches. However, tract 9813 is missing some patches in each filter:
HSC-G: 78 patches (missing:{0,0}, {0,8}, {8,0})
HSC-I: 79 patches (missing:{0,0}, {0,8})
HSC-R: 74 patches (missing:{0,0},{0,1},{0,7},{0,8},{1,8},{8,0},{8,8})
HSC-Z: 79 patches (missing:{0,0}, {0,8})
HSC-Y: 79 patches (missing: {0,0}, {0,8})
NB0921: 76 patches (missing: {0,0}, {0,7}, {0,8}, {1,8}, {8,8})
This closely mirrors the results shown in DM-11345 for all 9813 filters except for HSC-Z (that run was missing the {8,8} patch in addition to those mentioned above).
After checking to see which patches are included in deepCoadd-results/merged (created by multiBandDriver.py), I found that tracts 8766 and 8767 included all 81 patches, while tract 9813 included 79 patches (missing: {0,0}, {0,8}). The repeated visit/ccd IDs of "Removing visit" is known and expected; there would be one record from each patch if I understand correctly.
Note that the COSMOS visit list in DM-11345 is different from the one you use here. Some problematic visits have been removed for RC2.
Hsin-Fang Chiang
added a comment - The repeated visit/ccd IDs of "Removing visit" is known and expected; there would be one record from each patch if I understand correctly.
Note that the COSMOS visit list in DM-11345 is different from the one you use here. Some problematic visits have been removed for RC2. Ah, ok, that would make sense. In that case, is there anything else that I should include here before I close this ticket?
Samantha Thrush
added a comment - - edited Ah, ok, that would make sense. In that case, is there anything else that I should include here before I close this ticket? You can close the ticket if you have verified the run and are happy with the results.
Hsin-Fang Chiang
added a comment - You can close the ticket if you have verified the run and are happy with the results. Ok. I'm just running coaddDriver for Cosmos Y one more time just to confirm that the "removed visits" seen above isn't a complete fluke.
Samantha Thrush
added a comment - Ok. I'm just running coaddDriver for Cosmos Y one more time just to confirm that the "removed visits" seen above isn't a complete fluke. I've confirmed that the number of "Removing visits" is not a one-time fluke for CosmosY and am happy with my results.
Samantha Thrush
added a comment - I've confirmed that the number of "Removing visits" is not a one-time fluke for CosmosY and am happy with my results. 
All of the runs have finished successfully, as confirmed by Slurm logs and butler and all of the files are located at /project/hsc_rc/w_2018_03/
DM-13463.When I ran singleFrameDriver.py, I encountered 29 FATAL errors from the following ccd/visit pairs:
Wide: {visit: 7344, ccd: 67}, {visit:19468, ccd:69}, {visit:9708, ccd:99}, {visit:9736, ccd:67}, {visit:17738, ccd:69}, {visit:17750, ccd:58}, {visit:9868, ccd 76}, {visit:11582, ccd:76}, {visit:6478, ccd:99}, {visit:6528, ccd:24}, {visit:6528, ccd:67}, {visit:6528, ccd:59}
Cosmos: {visit:11698, ccd:68}, {visit:17934, ccd:1}, {visit:28376, ccd: 69}, {visit:28382, ccd:101}, {visit:28392, ccd:101}, {visit:28396, ccd:102}, {visit:28398, ccd:95}, {visit:28398, ccd:101}, {visit:28400, ccd:53}, {visit:28400, ccd:61}, {visit:28400, ccd:95}, {visit:28400, ccd:101}, {visit:28400, ccd:100}, {visit:23596, ccd:6}, {visit:280, ccd:22}, {visit:280, ccd:13}, {visit:280, ccd:103}
The errors encountered were:
File "src/PsfexPsf.cc", line 221, in virtual std::shared_ptr<lsst::afw::image::Image<double> > lsst::meas::extensions::psfex::PsfexPsf::_doComputeImage(const Point2D&, const lsst::afw::image::Color&, const Point2D&) const
Only spatial variation (ndim == 2) is supported; saw 0 {0}"
No other FATAL errors were found in the logs for makeSkyMap.py, mosaic.py, coaddDriver.py, and multiBandDriver.py. The log files can be found in /project/hsc_rc/w_2018_03/
DM-13463/logs. If you are interested in seeing the script used to run this reprocessing, it can be found in /project/hsc_rc/w_2018_03/DM-13463.Below is a table detailing the various pertinent information of the run, especially the node-hours used. To summarize this table quickly, this run used a total of 209.2774 node-hours.