[jira] Updated: (HIVE-1508) Add cleanup method to HiveHistory class
[ https://issues.apache.org/jira/browse/HIVE-1508?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Carl Steinbach updated HIVE-1508: - Component/s: Server Infrastructure Add cleanup method to HiveHistory class --- Key: HIVE-1508 URL: https://issues.apache.org/jira/browse/HIVE-1508 Project: Hive Issue Type: Bug Components: Metastore, Server Infrastructure Reporter: Anurag Phadke Assignee: Edward Capriolo Priority: Blocker Fix For: 0.7.0 Attachments: hive-1508-1-patch.txt Running hive server for long time 90 minutes results in too many open file-handles, eventually causing the server to crash as the server runs out of file handle. Actual bug as described by Carl Steinbach: the hive_job_log_* files are created by the HiveHistory class. This class creates a PrintWriter for writing to the file, but never closes the writer. It looks like we need to add a cleanup method to HiveHistory that closes the PrintWriter and does any other necessary cleanup. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.
[jira] Updated: (HIVE-1508) Add cleanup method to HiveHistory class
[ https://issues.apache.org/jira/browse/HIVE-1508?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Edward Capriolo updated HIVE-1508: -- Attachment: hive-1508-1-patch.txt Add cleanup method to HiveHistory class --- Key: HIVE-1508 URL: https://issues.apache.org/jira/browse/HIVE-1508 Project: Hive Issue Type: Bug Components: Metastore Reporter: Anurag Phadke Assignee: Edward Capriolo Priority: Blocker Fix For: 0.7.0 Attachments: hive-1508-1-patch.txt Running hive server for long time 90 minutes results in too many open file-handles, eventually causing the server to crash as the server runs out of file handle. Actual bug as described by Carl Steinbach: the hive_job_log_* files are created by the HiveHistory class. This class creates a PrintWriter for writing to the file, but never closes the writer. It looks like we need to add a cleanup method to HiveHistory that closes the PrintWriter and does any other necessary cleanup. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.
[jira] Updated: (HIVE-1508) Add cleanup method to HiveHistory class
[ https://issues.apache.org/jira/browse/HIVE-1508?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Edward Capriolo updated HIVE-1508: -- Status: Patch Available (was: Open) Did two things here. First random can generate the same result if called twice in the same millisecond. We guard against that with a while condition. Secondly leverage a finalizer to deal with clean up. Might be a nicer way, but not sure what is the best way to go about that. No easy way to unit test here, but you could run with lsof and determine if handles are ever closing. Add cleanup method to HiveHistory class --- Key: HIVE-1508 URL: https://issues.apache.org/jira/browse/HIVE-1508 Project: Hive Issue Type: Bug Components: Metastore Reporter: Anurag Phadke Assignee: Edward Capriolo Priority: Blocker Fix For: 0.7.0 Attachments: hive-1508-1-patch.txt Running hive server for long time 90 minutes results in too many open file-handles, eventually causing the server to crash as the server runs out of file handle. Actual bug as described by Carl Steinbach: the hive_job_log_* files are created by the HiveHistory class. This class creates a PrintWriter for writing to the file, but never closes the writer. It looks like we need to add a cleanup method to HiveHistory that closes the PrintWriter and does any other necessary cleanup. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.
