-
## Version
Same versions of the snaps on all cluster members.
```bash
root@m1:~# snap list
Name Version Rev Tracking Publisher Notes
core20 20230801 …
-
Each node currently caches tracking data in-memory.
The inclusion of Redis would allow us to track "is this cached" at a higher level across nodes.
At a basic level this would be a direct repla…
-
In setups with [Traefik Ingressroutes](https://doc.traefik.io/traefik/v2.2/routing/providers/kubernetes-crd/), a `LoadBalancer` is deployed by Traefik. When Cuber installs its own one next to it, is i…
-
#### Is your feature request related to a problem? Please describe
ArgoCD + deploy clusters with Crossplane is a popular choice of deployment model. We should make an example of it with EKS addons …
-
**Describe the bug**
The [comment][d] and the [code][a] suggest that if there are more than 1 worker, `preload_app` should be `true`, but it isn't.
**Puma config:**
The command line option is `-w…
-
It might be useful to have a subscription api for the cluster's current state. Perhaps a new ClusterDomainEvent could be created to make use of the existing subscription mechanism. The event can b…
-
Candidates for refactoring:
- [ ] logs-10-celery-error-1-minute-critical: This is currently tracking `"?\"ERROR/Worker\" ?\"ERROR/ForkPoolWorker\" ?\"WorkerLostError\""` found in cloudwatch `eks-cl…
-
- [ ] Packaging Uniformity(Use Automake and other Auto tools)
- [ ] Merge Glusterd2 spec file into `glusterfs.spec.in`
- [ ] Modify `tests/cluster.rc` to use `glusterd2` and fix the "*.t" files to…
-
I tried this by tracking occupants of nodes in the pathfinding. This just led to some really awkward pathfinding. Rather than try to go all the way around, they should wait until the optimal path is c…
-
`GivenKeysLoaderTask` prepares `Future`s of `PutFromLoadAll` operations which execute the actual batch loading. Once these futures are done, the record store is considered as loaded. However when a ne…