Arpit Agarwal created HADOOP-15450:
--------------------------------------
Summary: Avoid fsync storm triggered by DiskChecker and handle
disk full situation
Key: HADOOP-15450
URL: https://issues.apache.org/jira/browse/HADOOP-15450
Project: Hadoop Common
Issue Type: Bug
Reporter: Arpit Agarwal
Assignee: Arpit Agarwal
Fix disk checker issues reported by [~kihwal] in HADOOP-13738:
1. When space is low, the os returns ENOSPC. Instead simply stop writing, the
drive is marked bad and replication happens. This make cluster-wide space
problem worse. If the number of "failed" drives exceeds the DFIP limit, the
datanode shuts down.
1. There are non-hdfs users of DiskChecker, who use it proactively, not just on
failures. This was fine before, but now it incurs heavy I/O due to introduction
of fsync() in the code.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: common-dev-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-dev-h...@hadoop.apache.org
