-
Notifications
You must be signed in to change notification settings - Fork 3.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Peons put high load on zookeeper on disconnects #1970
Comments
nishantmonu51
added a commit
to metamx/druid
that referenced
this issue
Nov 24, 2015
…zk load apache#1970 - fixes apache#1970 - extract out segment handoff callbacks in SegmentHandoffNotifier which is responsible for tracking segment handoffs and doing callbacks when handoff is complete. - Overlord now maintains a view of segments in the cluster, this will affect the jam heap requirements for the overlord for large clusters. - realtime Index Tasks now use SegmentHandoffCheckAction to check for handoffs. - Realtime Nodes still use the old way of maintaining a FilteredServerView to check for handoffs. - Add tests for individual components fix broker test after merging
nishantmonu51
added a commit
to metamx/druid
that referenced
this issue
Dec 4, 2015
…dpoint for handoffs - fixes apache#1970 - extracted out segment handoff callbacks in SegmentHandoffNotifier which is responsible for tracking segment handoffs and doing callbacks when handoff is complete. - Coordinator now maintains a view of segments in the cluster, this will affect the jam heap requirements for the overlord for large clusters. realtime index task and nodes now use HTTP end points exposed by the coordinator to get serverView TODO: - Add more tests for individual components - Add docs Add tests and more loggings.
nishantmonu51
added a commit
to metamx/druid
that referenced
this issue
Dec 7, 2015
…dpoint for handoffs - fixes apache#1970 - extracted out segment handoff callbacks in SegmentHandoffNotifier which is responsible for tracking segment handoffs and doing callbacks when handoff is complete. - Coordinator now maintains a view of segments in the cluster, this will affect the jam heap requirements for the overlord for large clusters. realtime index task and nodes now use HTTP end points exposed by the coordinator to get serverView review comment fix realtime node guide injection review comments make test not rely on scheduled exec fix compilation fix import
nishantmonu51
added a commit
to metamx/druid
that referenced
this issue
Dec 8, 2015
…dpoint for handoffs - fixes apache#1970 - extracted out segment handoff callbacks in SegmentHandoffNotifier which is responsible for tracking segment handoffs and doing callbacks when handoff is complete. - Coordinator now maintains a view of segments in the cluster, this will affect the jam heap requirements for the overlord for large clusters. realtime index task and nodes now use HTTP end points exposed by the coordinator to get serverView review comment fix realtime node guide injection review comments make test not rely on scheduled exec fix compilation fix import review comment introduce immutableSegmentLoadInfo fix son reading remove unnecessary logging
nishantmonu51
added a commit
to metamx/druid
that referenced
this issue
Dec 9, 2015
…dpoint for handoffs - fixes apache#1970 - extracted out segment handoff callbacks in SegmentHandoffNotifier which is responsible for tracking segment handoffs and doing callbacks when handoff is complete. - Coordinator now maintains a view of segments in the cluster, this will affect the jam heap requirements for the overlord for large clusters. realtime index task and nodes now use HTTP end points exposed by the coordinator to get serverView review comment fix realtime node guide injection review comments make test not rely on scheduled exec fix compilation fix import review comment introduce immutableSegmentLoadInfo fix son reading remove unnecessary logging
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Realtime Index tasks need to know when the segments is handed over to the historical nodes.
To get this information realtime index task keeps a FilteredBrokerserverView and gets data from zookeeper for all the segments in the cluster. In case of ZK disconnects, each task will reconnect to zookeeper and starts reading the distribution of segments in the cluster which puts a high load on zookeeper as the number of tasks starts to grow.
One solution to handle this is to make the overlord responsible for keeping the state segments in cluster and the realtime nodes will interact with the overlord to know when a segment has been handed off to another historical node.
The text was updated successfully, but these errors were encountered: