Local cache out of sync for docker runner - gitlab-runner 11.8.0 on arch linux
Cache created by a stage is not correctly read by later stages - possibly after a clean runners cache button being pressed. This can be seen in pipeline: https://gitlab.com/LinkiTools/racket/pipelines/51600198
Steps to reproduce
prepare-cache:qemu builds qemu 3.1.0 and stores it in the cache for later jobs
test:<arch>. All was working find until I started cleaning the cache with the 'clean runners cache' button in the pipelines view.
prepare-cache:qemu that has cache settings:
variables: INSTALL_DIR: $CI_PROJECT_DIR/install cache: key: qemu-3.1.0 paths: - $INSTALL_DIR
After the clean job
Creating cache qemu-3.1.0-1... /builds/LinkiTools/racket/install: found 112 matching files Archive is up to date! Created cache
For some reason there's now a
-1 suffix to the cache key. In any case, I didn't expect later stages to fail but they do. Job
test:arm64 for example shows:
Running with gitlab-runner 11.8.0 (4745a6f3) on andromeda-racket yyn-fYs4 Using Docker executor with image ubuntu:18.04 ... Pulling docker image ubuntu:18.04 ... Using docker image sha256:94e814e2efa8845d95b2112d54497fbad173e45121ce9255b93401392f538499 for ubuntu:18.04 ... Running on runner-yyn-fYs4-project-8928499-concurrent-0 via andromeda... Fetching changes... HEAD is now at 29551644d0 Show current directory Checking out 29551644 as pmatos-qemu-testing... Skipping Git submodules setup Checking cache for qemu-3.1.0-2... No URL provided, cache will not be downloaded from shared cache server. Instead a local version of cache will be extracted. Successfully extracted cache
This fails because there's no such cache. Cache is
Please see: https://gitlab.com/LinkiTools/racket/
What is the current bug behavior?
test:<arch> fail because they look for wrong cache.
What is the expected correct behavior?
Cache is found and extracted correctly.
Relevant logs and/or screenshots
Happy to add something else here if you need me to.
Output of checks
This bug happens on GitLab.com