My paths were different, and it was .tabledesc rather than .tableinfo, but I got past the problem. Now the restore_snapshot seems to be hung, and I’m seeing many warnings like the following in the logs:
master.RegionStates: Failed to open/close 24635905eba622ed4911a498b7848caa on foo.com,60020,1395321350441, set to FAILED_OPEN On Mar 21, 2014, at 12:40 PM, Matteo Bertozzi <[email protected]> wrote: > sorry forgot the namespace: 96 path is > /hbase/archive/data/default/TABLE_NAME > > Matteo > > > > On Fri, Mar 21, 2014 at 4:38 PM, Matteo Bertozzi > <[email protected]>wrote: > >> The directory layout is changed, but easy to fix by hand. >> >> 94 path is /hbase/.archive/TABLE_NAME -> 96 path is >> /hbase/archive/data/TABLE_NAME >> 94 table info is in .hbase-snapshot/NAME/.tableinfo.xyz -> 96 table info >> is in .hbase-snapshot/NAME/.tableinfo/tableinfo.xyz >> >> Matteo >> >> >> >> On Fri, Mar 21, 2014 at 4:27 PM, Brian Jeltema < >> [email protected]> wrote: >> >>> Exporting across versions was why I tried webhdfs. I have a cluster >>> running HBase 0.94 and wanted to export a >>> table to a different cluster running HBase 0.96. I got the export to >>> work, but attempting to >>> do a restore_snapshot results in: >>> >>> TableInfoMissingException: No table descriptor file >>> >>> so I assume snapshot are not compatible between versions. Is there any >>> way to copy the >>> table from the old cluster to the new one? >>> >>> Brian >>> >>>> Also Matteo, just like distcp, one advantage of this (using webhdfs >>> while >>>> copying) could also be that even if the versions are not same, we can >>> still >>>> copy? >>>> >>>> Regards, >>>> Shahab >>>> >>>> >>>> On Fri, Mar 21, 2014 at 8:14 AM, Matteo Bertozzi < >>> [email protected]>wrote: >>>> >>>>> ExportSnapshot uses the FileSystem API >>>>> so you'll probably be able to say: -copy-to: webhdfs://host/path >>>>> >>>>> Matteo >>>>> >>>>> >>>>> >>>>> On Fri, Mar 21, 2014 at 12:09 PM, Brian Jeltema < >>>>> [email protected]> wrote: >>>>> >>>>>> Is it possible to use webhdfs to export a snapshot to another >>> cluster? If >>>>>> so, >>>>>> what would the command look like? >>>>>> >>>>>> TIA >>>>>> >>>>>> Brian >>>>>> >>>>> >>> >>> >>
