Bug 927420

Summary:	GlusterFileSystem.append() throws IOException if called
Product:	[Community] GlusterFS	Reporter:	Steve Watt <swatt>
Component:	gluster-hadoop	Assignee:	Bradley Childs <bchilds>
Status:	CLOSED EOL	QA Contact:	Martin Kudlej <mkudlej>
Severity:	high	Docs Contact:
Priority:	medium
Version:	mainline	CC:	bugs, chrisw, eboyd, matt, mkudlej, rhs-bugs, vbellur
Target Milestone:	---	Keywords:	Triaged
Target Release:	---
Hardware:	x86_64
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2016-02-01 16:15:32 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:
Bug Blocks:	947153, 1057253

Description Steve Watt 2013-03-25 21:08:59 UTC

Description of problem:

append is implemented in the DistributedFileSystem class. I'm unsure why this is not implemented in GlusterFileSystem at all. Given that we are POSIX compliant I would imagine this behavior is relatively simple to implement.

Comment 2 Bradley Childs 2013-04-26 18:34:54 UTC

Per the apache HCFS discussion, append(..) is a MAY implement for filesystems. 

append(..) is generally avoided all together as it raises concurrency issues that aren't well defined in HCFS/HDFS.  The code to implement isn't hard, but i'd suggest avoiding this as long as possible. At the most defined state this would be a very slow right path, and at the least defined state it could cause consistency problems. 

The behavior or append(..) operations to already opened files, or when the append(..) data becomes consistent isn't well defined for HCFS. 

Ref:
https://issues.apache.org/jira/browse/HADOOP-9371

Steve, I'll re-assign to you for decision close or size & plan if so desired.

Comment 3 Steve Watt 2013-05-01 21:45:04 UTC

I think there is a possibility that FS Append semantics are required by HBase for its Write Ahead Log (WAL) and we match expected FileSystem semantics for the standard clients of the Hadoop FileSystem. Can you verify that the append semantics are not required by HBase?

Comment 4 Bradley Childs 2013-08-20 15:48:16 UTC

code re-written as of 2.0. 

GlusterFileSystem extends FilterFileSystem
GlusterVolume extends RawLocalFileSystem.

the append behavior is inherited (and supported) in RawLocalFileSystem.  

this bug can be closed.

Comment 6 Steve Watt 2016-02-01 16:15:32 UTC

This solution is no longer available from Red Hat.