[HBASE-8722] Specify and document the FileSystem level behaviors required by HBase

• Type: Umbrella
• Status: Closed
• Priority: Major
• Resolution: Not A Problem
• Assignee: Unassigned
• Reporter: Andrew Kyle Purtell
• Created: 10/Jun/13 17:18
• Updated: 16/Jun/22 17:08
• Resolved: 12/Jun/22 00:00

Description

HADOOP-9361 is an umbrella for an effort to strictly define the expected behavior of filesystem APIs, and then write tests to verify compliance. Currently we implicitly require some HDFS semantics that may not be available in other filesystems or data storage options. We should specify and document the filesystem level behaviors expected by HBase so they are visible as part of this effort. Then filesystem providers can know what is required for HBase application support through spec and standalone tests and won't need to bootstrap a full HBase installation, which can be a complex and error-prone undertaking if the semantics of the underlying storage are incompatible.

Some areas that come to mind are: close-to-open, pread, lease recovery, flush/sync, atomic renames.

Issue Links

• is related to: HADOOP-9361 Strictly define the expected behavior of filesystem APIs and write tests to verify compliance