[ http://issues.apache.org/jira/browse/HADOOP-334?page=comments#action_12447579 ] Sameer Paranjpye commented on HADOOP-334: -----------------------------------------
Dhruba: Yes, we would get correct semantics if the image and edits overlap. While not strictly necessary, we could also add timestamps to entries in the image and edits for validation. Suppose the edits file has the following transactions: - x: add file /a/b/c, time=t0 - y: remove file /b/c, time=t1 - z: rename directory /a --> /d, time=t2 which all overlap with the image. In this case, when the image and edits are being merged the following would happen for each transaction: x: the image already has file /a/b/c, assert that directory /a/b has timestamp t3 >= t0, we're consistent with the edits, move on y: the image does not have the file /b/c, assert that directory /b has timestamp t4 >= t1, we're consistent with the image, move on z: the image already has the directory /d, assert that directory /d has timestamp t5 >= t2, we're consistent with the edits, move on this sort of logic can be followed for all transactions. So we don't need a transaction number, a timestamp will suffice. > Redesign the dfs namespace datastructures to be copy on write > ------------------------------------------------------------- > > Key: HADOOP-334 > URL: http://issues.apache.org/jira/browse/HADOOP-334 > Project: Hadoop > Issue Type: Improvement > Components: dfs > Affects Versions: 0.4.0 > Reporter: Owen O'Malley > Assigned To: Konstantin Shvachko > > The namespace datastructures should be copy on write so that the namespace > does not need to be completely locked down from user changes while the > checkpoint is being made. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: http://issues.apache.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira
