we updated our artifactoy to the newest version 5.10.3 to support the pypi update.
after this update we experience some really slow pip install --upgrade times.
pip install is running fine.
I was able to pinpoint this slower runs to the local pypi repositiory on the artifactory.
The official pypi and the cached remote pypi on the artifactory were fast.
But the virtual pypi (remote und local pypi) and if i only add the local pypi via --extra-index-url were both slow
Some versions, maybe we have some issues there:
Run with official Pypi only:
pip install -U -r requirements.txt 1.21s user 0.08s system 99% cpu 1.296 total
Run with local pypi as extra-index-url:
pip install -U -r requirements.txt 1.51s user 0.09s system 8% cpu 19.658 total
And with the virtual pypi:
pip install -U -r requirements.txt 1.57s user 0.10s system 4% cpu 38.573 total
Unfortunately i don't have and time before the update.
some ideas?
The issue was fixed in the Artifactory Release 5.10.4
The corresponding Ticket is https://www.jfrog.com/jira/browse/RTFACT-16513
Degradation happens on Derby DB in case node_props table has many entries, due to a changed DB query introduced at RTFACT-15433
After the upgrade, the pypi repos will be fast again.