Hi, You have two options to disable anticompaction when running full repair:
- add the list of DCs using the --dc flag (even if there's just a single DC in your cluster) - Use subrange repair, which is done by tools such as Reaper (it can be challenging to do it yourself on a vnode cluster). You'll have to mark the sstables which are marked as repaired back to unrepaired state. This operation requires to stop one node at a time, using the sstablerepairedset tool (check the official Cassandra docs for more info). FTR, Cassandra 4.0 will not perform anticompaction anymore on full repairs. Cheers, Alex Le lun. 9 nov. 2020 à 05:57, Pushpendra Rajpoot < pushpendra.nh.rajp...@gmail.com> a écrit : > Hi Team, > > In Cassandra 3.x, Anti-compaction is performed after repair (incremental > or full). Repair does not have any way to bypass anti-compaction (if not > running sub range repair with -st & -et). Here is a jira ticket. > > https://issues.apache.org/jira/browse/CASSANDRA-11511 > > I am facing 100% disk utilization when running repair -pr after upgrading > Cassandra from 2.1.16 to 3.11.2. I have a 2TB disk & 1.35TB is used. > > There are multiple keyspaces and each keyspace is having multiple tables. > One of the sstable has huge data. Here is the details of it : Keyspace, > tabel & SStable have 1.3TB, 730GB, 270 GB of data. > > I have following questions : > > 1. Is there any way to disable anti-compaction after repair is > completed ? > 2. After what stage, anti-compaction is performed after repair ? > 3. Any other suggestions? > > Regards, > Pushpendra >
