- 19 Apr, 2020 9 commits
-
-
Daniel Stone authored
Maintaining the cloud-init YAML by hand was getting pretty old, especially as we needed to inline some things like the docker-gc script. Rather than doing that, rewrite our cloud-init script as a Python generator, which will allow us to more easily parameterise in order to support things like different Docker GC strategies, different architectures and disk layouts, etc - and not have to manually inline our scripts. This is currently implemented as a giant dict which matches the output YAML, to ease diffing, but will become more programmatic over later commits. Signed-off-by:
Daniel Stone <daniels@collabora.com>
-
I don't know whether or not this is actually a good idea. On one hand, if we exclude them, then we cut runtime down, which means that our runners are more efficiently utilised, which means that we need fewer runners to serve our jobs, which means we have a better chance of hitting local cache. On the other hand, the upstream images don't cost us anything to pull, and are competing with space for local images. I think the best thing to do would be to see the effect after a week or so, and whether or not we have space for them without jettisoning lots of local images. Signed-off-by:
Daniel Stone <daniels@collabora.com>
-
Based on just looking at 'docker images' output. Signed-off-by:
Daniel Stone <daniels@collabora.com>
-
Benjamin Tissoires authored
-
Daniel Stone authored
Signed-off-by:
Daniel Stone <daniels@collabora.com>
-
Daniel Stone authored
Signed-off-by:
Daniel Stone <daniels@collabora.com>
-
Daniel Stone authored
Ignore local mounts like the global cache. Signed-off-by:
Daniel Stone <daniels@collabora.com>
-
Daniel Stone authored
Signed-off-by:
Daniel Stone <daniels@collabora.com>
-
- 17 Apr, 2020 1 commit
-
-
Benjamin Tissoires authored
-
- 16 Apr, 2020 2 commits
-
-
Benjamin Tissoires authored
-
Benjamin Tissoires authored
The controller all the time removes and adds the following: ``` ssl_trusted_certificate /etc/ingress-controller/ssl/default-gitlab-pages-tls-full-chain.pem; ssl_stapling on; ssl_stapling_verify on; ```
-
- 15 Apr, 2020 2 commits
-
-
Benjamin Tissoires authored
-
Benjamin Tissoires authored
It seems the CN doesn't need this as there is a job creating the secrets on upgrade. But for us, we need it or the ssh port will not be forwarded. Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@gmail.com>
-
- 14 Apr, 2020 1 commit
-
-
Benjamin Tissoires authored
For the new charts to work, we need to namespace all of the config files because we are using the root of the directory, not just `./charts/gitlab-omnibus`. This allows to be more compatible with the gitlab native kubernetes chart and should allow us to move forward. Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@gmail.com>
-
- 10 Apr, 2020 1 commit
-
-
Benjamin Tissoires authored
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
- 08 Apr, 2020 19 commits
-
-
Benjamin Tissoires authored
To simplify the calls to get_correct_iso_format() Then: `%s/datetime.datetime.fromisoformat(get_correct_iso_format(\(.*\)))/docker_fromisoformat(\1)/g` Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
Some images are not part of the db, either the db is not up to date, either they were pulled manually. Ignore them. Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
so that we do not remove objects that have been created/used less than a day ago. Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
Now that the history is offloaded to the gitlab server, we can start we accurate data. Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
This should remove the complex need to listen to events, not have history and whatnot. Right now it is very simple, it counts since the beginning of the available files in fdo-containers-usage. Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
problem is that the stored docker events are not that many, and we mostly only get events from gitlab/gitlab-runner-helper Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
We want to listen to events, so clearing up our state at every run will make this moot. Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
this makes easier to sort the list of last used images Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
otherwise: File "/usr/lib/python3/dist-packages/docker/errors.py", line 31, in create_api_error_from_http_exception raise cls(e, response=response, explanation=explanation) docker.errors.NotFound: 404 Client Error: Not Found ("No such container: 03c5217472db4bc979a167371262ce56e93dc2a9fd7e19791fa59ecf4cb0ef24") Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
We need to enforce working with UTC, otherwise we won't get accurate expired time stamps. Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
Benjamin Tissoires authored
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@redhat.com>
-
- 07 Apr, 2020 1 commit
-
-
Emma Anholt authored
As a reminder to myself in the future, this was generated with: kubectl get pods --namespace marge-bot -o yaml > marge-mesa-config.yaml
-
- 28 Mar, 2020 4 commits
-
-
allows to run the script just once or in a loop Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@gmail.com>
-
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@gmail.com>
-
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@gmail.com>
-
Signed-off-by:
Benjamin Tissoires <benjamin.tissoires@gmail.com>
-