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
In order to enable locality-aware logs configuration, we need to support configuring the locality for nodes. Based on this information, the cluster controller and other components can make decisions on where to place loglet replicas and which nodes to replicate data to.
One idea could be to allow users to specify a set of key value pairs at startup of the form --locality=region=us-west-1,zone=sf,rack=1 which describe the location of the node. This information could be recorded in the NodesConfiguration to make it accessible to the various components that need this information. This approach would also allow to specify other labels that could be considered when placing loglets. If the locality value starts with the most inclusive scope (region, then az, then something else), then we could pick nodes such that we first try to find nodes with different values for the most inclusive scope, then continue with the second most and so on (looking for nodes with different prefix values and expanding the prefix if needed).
The text was updated successfully, but these errors were encountered:
In order to enable locality-aware logs configuration, we need to support configuring the locality for nodes. Based on this information, the cluster controller and other components can make decisions on where to place loglet replicas and which nodes to replicate data to.
One idea could be to allow users to specify a set of key value pairs at startup of the form
--locality=region=us-west-1,zone=sf,rack=1
which describe the location of the node. This information could be recorded in theNodesConfiguration
to make it accessible to the various components that need this information. This approach would also allow to specify other labels that could be considered when placing loglets. If thelocality
value starts with the most inclusive scope (region, then az, then something else), then we could pick nodes such that we first try to find nodes with different values for the most inclusive scope, then continue with the second most and so on (looking for nodes with different prefix values and expanding the prefix if needed).The text was updated successfully, but these errors were encountered: