There has been some recent discussions on http://issues.apache.org/jira/browse/HADOOP-1700. It is likely that a single-writer append design and implementation will come soon.
This jira issue is supposed to capture the expectations and requirements of "append", the type of cache coherency that is required, the persistency guarantees of incremental updates. etc.etc. If any of us know of applications that would use "append" as soon as it is ready, it would be really helpful if the requirements from that application can be added to this JIRA. It would help in making the right design decisions. Thanks, dhruba -----Original Message----- From: Ted Dunning [mailto:[EMAIL PROTECTED] Sent: Friday, August 31, 2007 9:22 AM To: [email protected] Subject: Re: Using Map/Reduce without HDFS? The priority is already listed as major. Nobody is assigned the bug, however. Presumably this won't be the kind of thing an outsider could do easily. On 8/31/07 8:48 AM, "mfc" <[EMAIL PROTECTED]> wrote: > > Not having the ability to append to a file in hadoop seems to be rather > limiting. > How can this get higher on the priority list? Even just a single appender. > > Thanks > > > Ted Dunning-3 wrote: >> >> >> Or lots and lots of little files. >> >> None of these is very attractive. >> >>
