Closed vg-github closed 7 years ago
hi @vladimirghetau , it seems you're running gtfs2ed
to read an ntfs, but it is done to read a gtfs.
you should use fusio2ed
to read the ntfs .
Hope this helps :smile:
So, my new understanding is this:
gtfs2ed
fusio2ed
osm2ed
?The problem now is, I did try to import the NTFS using fusio2ed and I get this:
INFO - add_feed_info, Key :ntfs_version Value :0.3 not imported. INFO - add_feed_info, Key :fusio_url Value :http://vip-fusio-ihm.UK.prod.canaltp.fr/ not imported. INFO - add_feed_info, Key :fusio_version Value :1.10.85.204 not imported. INFO - date de production: 2016-Sep-03 - 2017-Jan-03 INFO - Reading geometries INFO - Nb shapes: 0 INFO - default agency tz Europe/London -> GMT WARN - Impossible to parse the co2_emission for CheckIn Boarding WARN - Impossible to parse the co2_emission for CheckOut Landing WARN - Impossible to read /srv/ed/data/uk_ntfs/line_groups.txt WARN - Impossible to read /srv/ed/data/uk_ntfs/line_group_links.txt WARN - Impossible to read /srv/ed/data/uk_ntfs/odt_conditions.txt INFO - reading stop times Killed
Also, when importing a GTFS file, I get something similar:
/usr/bin/gtfs2ed -i "/srv/ed/data/uk_gtfs" --connection-string="host=localhost user=navitia dbname=navitia password=navitia" WARN - Impossible to read /srv/ed/data/uk_gtfs/feed_info.txt INFO - Unable to find production date in add_feed_info. INFO - date de production: 2016-Sep-03 - 2017-Jan-03 WARN - Impossible to read /srv/ed/data/uk_gtfs/shapes.txt INFO - default agency tz Europe/London -> GMT INFO - reading stop times Killed
What could be the problem? I couldn't find any log either to explain why the execution was killed.
hum the killed
is strange. Do you have enough memory ?
Do you have enough RAM? It's looks like the OOM killer. In our production the kraken for uk dataset use 14go of ram, most of these are from the stop times.
You are right osm2ed
is used for importing osm data from a pbf file, after having imported all your data you will need to run ed2nav
to generate a data.nav.lz4
that will be loaded by kraken.
For most of your usage, I guess you can get some help in https://github.com/CanalTP/navitia/blob/9bd3f700197234d752874a39b23fd7d876e9504b/documentation/vagrant/vagrant.md
And loading the GTFS itself using gtfs2ed might be a little cheaper in RAM, but you might also loose some data quality (transfers for example)
@pbougue - is there an alternative for gtfs2ed that loads data without loosing data quality?
Hmm, I don't know precisely what data adjustment we do on top of uk GTFS. I'd say we at least add transfers and a stop_area creation from close stop_points... And I just checked that we have them in the GTFS provided on OpenDataSoft platform. So it should be pretty much ok, comparing to the NTFS.
But you will still encounter a big RAM need when loading street network and transportation datas in kraken for the real journey planning.
I am trying to import the UK GTFS data from the Navitia OpenDataSoft:
https://navitia.opendatasoft.com/explore/dataset/uk/table/?sort=type_file
I get many errors, depending on the dataset I'm importing, and it seems each dataset has its own problems. Is there a way to import these datasets in a pre-validation stage first, I'm not sure what is Navitia expecting from each dataset, but it seems the might need a bit of enrichment/standardisation first.
For example: the UK gtfs.zip file, the error I'm getting from ED is:
For nfts.zip I get :
For the last zip file, uk_ouk_uk_national.zip:
Any feedback?