I ran test suite using https://github.com/francisliu/hbase_namespace/tree/core_8408_rebase_1 last week and got several test failures. See tail of this email.
I think namespace feature still needs some polishing. Francis and people interested in this feature can continue to work on the git branch so that all major issues (including migration) are ironed out. Cheers Failed tests: testCompactionRecordDoesntBlockRolling(org.apache.hadoop.hbase.regionserver.wal.TestLogRolling): Should have WAL; one table is not flushed expected:<1> but was:<2> testLogRollOnDatanodeDeath(org.apache.hadoop.hbase.regionserver.wal.TestLogRolling): New log file should have the default replication instead of 1 testFavoredNodesPresentForRoundRobinAssignment(org.apache.hadoop.hbase.master.TestRegionPlacement) Tests in error: org.apache.hadoop.hbase.catalog.TestMetaMigrationConvertingToPB: java.net.URISyntaxException: Relative path in absolute URI: hbase:meta testCorrectnessWhenMasterFailOver(org.apache.hadoop.hbase.regionserver.TestRSKilledWhenMasterInitializing): test timed out after 120000 milliseconds testRunShellTests(org.apache.hadoop.hbase.client.TestShell): (RuntimeError) Shell unit tests failed. Check output file for details. On Mon, Jul 29, 2013 at 6:46 PM, Enis Söztutar <enis....@gmail.com> wrote: > For namespaces, I was sidetracked for most of last couple of weeks and this > week as well. I did play with the code a bit and did a couple of > iterations, but did not get the chance to look into the latest patch. I > know Stack also looked at it. Unless there are other reviewers, I guess it > won't make into 0.95.2. This has been said many times before though that it > is very hard to bring this patch as it is after 0.96 goes out, because > rolling restarts will be broken. > > I am not sure why there is not much interest in this though. It adds > namespaces, and fixes the system tables problem. Let us reach a decision on > this either way and move on I guess. > > Enis > > > On Mon, Jul 29, 2013 at 11:50 AM, Andrew Purtell <apurt...@apache.org > >wrote: > > > On Mon, Jul 29, 2013 at 11:43 AM, Andrew Purtell <apurt...@apache.org > > >wrote: > > > > > On Mon, Jul 29, 2013 at 10:51 AM, Elliott Clark <ecl...@apache.org> > > wrote: > > > > > >> On Mon, Jul 29, 2013 at 10:26 AM, Stack <st...@duboce.net> wrote: > > >> > Main objection is that the included v3 is not a 'green field'/redo > so > > >> lets > > >> > not call it that > > >> > > >> We added minor version into hfile versioning (Broke things while doing > > >> it), so maybe it's better to call it 2.5 ? > > >> > > > > > > Works for me if that's the preference. Key is being able to switch on > > > something that might introduce new overheads only if tags are desired > on > > > account of some use case that depends on them. > > > > > > > The consequence is we will have less freedom to make new classes for tag > > specific handling. Granted the work to date has bled through, but that > > could change. So another option is copy / extend more code until v3 > meets a > > consensus standard of greenfield. > > > > -- > > Best regards, > > > > - Andy > > > > Problems worthy of attack prove their worth by hitting back. - Piet Hein > > (via Tom White) > > >
