We are running into too many out of disk space situations with OWS on
the main disk.
This way, we can reuse the Gerrit disk for all that data, which
hopefully, is quite shared with Gerrit.
Signed-off-by: Raito Bezarius <masterancpp@gmail.com>
This help us getting rid of useless traffic by crawlers.
It is enabled for gerrit01 which is suffering the most from this.
Signed-off-by: Raito Bezarius <masterancpp@gmail.com>
Lix may have its own secrets and we want to maintain a certain
generalization level on the NixOS modules, so we can decorrelate which
secret we select dynamically by having a simple tenancy hierarchy
system.
This unfortunately requires to rewrite all call sites with a floral
prefix until we migrate them to the simple internal secret module which
is aware of this.
Signed-off-by: Raito Bezarius <masterancpp@gmail.com>
Still not working due to "unimplemented: error 404 not found" at push
time, but it's really unclear now why this occur.
Signed-off-by: Raito Bezarius <masterancpp@gmail.com>
We had some issues in the past with too many packfiles and timeout
during transfers, let's try to provide a bit of relief in bad scenarios.
Signed-off-by: Raito Bezarius <masterancpp@gmail.com>
The local refspec was weird and exploiting a edge case for the nixpkgs
jobs where local and from were the same.
We are more explicit now, which fixes the sandbox jobs.
Signed-off-by: Raito Bezarius <masterancpp@gmail.com>
It's a modified version of @puck's Lix buildbot checks for
gerrit.lix.systems with a slight generalization in the configuration for
many repositories.
Signed-off-by: Raito Bezarius <masterancpp@gmail.com>
* flip off proxy_buffering again
* enable REVWALK_USE_PRIORITY_QUEUE
* enable delta compression, because that's not a bottleneck and it's
nicer on bandwidth
