Bill Farner created AURORA-122: ---------------------------------- Summary: Introduce a task store secondary index on slave host Key: AURORA-122 URL: https://issues.apache.org/jira/browse/AURORA-122 Project: Aurora Issue Type: Story Components: Scheduler Reporter: Bill Farner Assignee: Bill Farner
We've observed the {{resourceOffers}} callback taking longer than desired in large clusters. Timing has indicated that they take as long as 100ms regularly, with occasional calls taking as long as a minute. We've narrowed down the majority of time spent to be performing expensive task store queries by slave host (in {{GcExecutorLauncher}}. When lots of hosts and tasks are present, these calls should be considerably faster if we maintain a index on slave host to avoid full task store scans. -- This message was sent by Atlassian JIRA (v6.1.5#6160)