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
The sharding mechanism requires the shard lease names (i.e., controller instance identities) to be unique in a ring.
If there are multiple shard leases in different namespaces with the same name belonging to the same ring, this is likely a misconfiguration.
Typically, all controller instances of the same ring would reside in the same namespace and, thus, have non-conflicting names.
Alternatively, controller instances could belong to different rings and might use conflicting names with instances of other rings.
If – for whatever reason – there are multiple shards of the same ring with the same name in different namespaces, the sharder should handle this.
E.g., by adding an error condition to the ControllerRing status or annotating the Lease objects.
Instead of performing undesired rebalancing, the sharder should halt the sharding mechanism entirely until the conflicting names have been resolved.
The documentation should be updated to mention this requirement for controllers implementing sharding.
If someone wants to run controller instances of the same ring in different namespaces, they can do so by using unique shard names.
Why is this needed:
Having conflicting shard lease names sounds like an edge case to me and I don't think it restricts any reasonable scenarios.
Still, we should confusion and undesired sharder actions when running into such a situation.
The text was updated successfully, but these errors were encountered:
What would you like to be added:
The sharding mechanism requires the shard lease names (i.e., controller instance identities) to be unique in a ring.
If there are multiple shard leases in different namespaces with the same name belonging to the same ring, this is likely a misconfiguration.
Typically, all controller instances of the same ring would reside in the same namespace and, thus, have non-conflicting names.
Alternatively, controller instances could belong to different rings and might use conflicting names with instances of other rings.
If – for whatever reason – there are multiple shards of the same ring with the same name in different namespaces, the sharder should handle this.
E.g., by adding an error condition to the
ControllerRing
status or annotating theLease
objects.Instead of performing undesired rebalancing, the sharder should halt the sharding mechanism entirely until the conflicting names have been resolved.
The documentation should be updated to mention this requirement for controllers implementing sharding.
If someone wants to run controller instances of the same ring in different namespaces, they can do so by using unique shard names.
Why is this needed:
Having conflicting shard lease names sounds like an edge case to me and I don't think it restricts any reasonable scenarios.
Still, we should confusion and undesired sharder actions when running into such a situation.
The text was updated successfully, but these errors were encountered: