Description
A node restart currently iterates through each core and publishes an item to the Overseer queue to mark that core as 'down'. This is inefficient if each node has many cores and causes overseer to be overwhelmed with requests. We can publish a single 'down' status for the entire node and have the overseer do the rest.
Attachments
Attachments
Issue Links
- is duplicated by
-
SOLR-8616 Consider using multiple threads for publishCoresAsDown.
-
- Resolved
-
- relates to
-
SOLR-10277 On 'downnode', lots of wasteful mutations are done to ZK
-
- Closed
-
-
SOLR-8720 ZkController#publishAndWaitForDownStates should use #publishNodeAsDown.
-
- Resolved
-