I have noticed a build of the Jenkins job pywikibot-core-tox that spent a lot of time syncing the cache:
00:00:01.240 Syncing... 00:13:01.139 00:13:01.139 Done
That is 13 minutes to restore pre-commit caches.
The duration has kept growing, here over a year (Grafana):
The namespace is CASTOR_NAMESPACE="pywikibot-core/master/pywikibot-core-tox", looking at the Castor instance:
ssh integration-castor05.integration.eqiad1.wikimedia.cloud sudo du -m -d1 /srv/castor/pywikibot-core/master/pywikibot-core-tox 1182 /srv/castor/pywikibot-core/master/pywikibot-core-tox/pip 60 /srv/castor/pywikibot-core/master/pywikibot-core-tox/virtualenv 11446 /srv/castor/pywikibot-core/master/pywikibot-core-tox/pre-commit 12688 /srv/castor/pywikibot-core/master/pywikibot-core-tox
pre-commit stores caches there that keeps accumulating. Some of the cache directories take 189MB, they a couple pyenv:
94 ./py_env-python3.10 95 ./py_env-python3.13
I don't think we should store them, or at least they should not be accumulating them? For the tox/pip part, we only save and restore the pip cache, the venv are reinstalled from scratch on each build.
