Hi Jean, Our concern is the repair, in 3.0.7 repairs inc are by default. Then it > means that once we do the upgrade to 3.0.7 we must follow the migration > process of repairs inc for all our data in order to mark the sstables as > repaired ?
If you do not feel ready for incremental repairs, just adding the '-full' option to your 'nodetool repair' command should be enough to continue repairing as you currently are once using 3.0.7. 'nodetool repair -inc' becomes the default 'nodetool repair' indeed, but you're not forced to use incremental repair because you will be using 3.0.7, that's why '-full' option was added. You did the hardest part in noticing this change and start wondering about it. C*heers, ----------------------- Alain Rodriguez - @arodream - al...@thelastpickle.com France The Last Pickle - Apache Cassandra Consulting http://www.thelastpickle.com 2016-09-12 13:55 GMT+02:00 Paulo Motta <pauloricard...@gmail.com>: > Migration procedure is no longer required for incremental repair as of > 2.1.4 since CASSANDRA-8004, which was the reason why the migration > procedure was required for LCS before. The migration procedure is only > useful now to skip validation on already repaired sstables in the first > incremental repair run by marking them as repaired before running the first > incremental repair, otherwise anti-compaction will mark them as repaired in > the first run since CASSANDRA-7586 on 2.2+. > > 2016-09-06 5:26 GMT-03:00 Jean Carlo <jean.jeancar...@gmail.com>: > >> Hello guys >> >> We are planing to upgrade cassandra soon to the version 3.0.7 from >> 2.1.14. Our concern is the repair, in 3.0.7 repairs inc are by default. >> Then it means that once we do the upgrade to 3.0.7 we must follow the >> migration process of repairs inc for all our data in order to mark the >> sstables as repaired ? or we can just run directly the repair command >> without need to mark the sstables previously? >> >> My first test with ccm tells me that we don't need to mark the sstables >> because the repair in the 3.0.7 do it for you, but I want to ask if someone >> has done this migration and confirm my assumption >> >> Best regards. >> >> Jean Carlo >> >> "The best way to predict the future is to invent it" Alan Kay >> > >