You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
That would be a bit more work as chaoskube currently works solely on the Pod list.
One would have to use the ownerReferences to find its ReplicaSet and from there the current number of siblings.
Another but more hackier way would be to take the labels of the potential target Pod and make sure there's at least one other Pod matching the same labels before killing it. (Other Pods with the same labels are likely members of the same Deployment I would assume.)
An alternative would be to continue work on respecting PodDisruptionBudgets (PDBs)(#13) in which case you could "protect" your single Pod. However, that would also effect all other components working on PDBs such as controlled cluster updates which might not like that. (PDBs for single replica Deployments aren't really good practice.)
I think this makes a lot of sense in order to start using chaoskube without unnecessary interruptions.
There are applications that aren't high-available and if their single replica dies they are just gone for a moment. If you're fine with that and still want to use chaoskube on all of your other stuff, there should be an option to do just that.
Let's go with:
One would have to use the ownerReferences to find its ReplicaSet and from there the current number of siblings.
Furthermore, anything that isn't reconciled by a controller (e.g. kubectl run -i -t busybox --image=busybox --restart=Never) should also be excluded in this mode.
I'd like the ability to exclude apps with only 1 replica; any ideas? Would require a change I presume...
The text was updated successfully, but these errors were encountered: