If you are not deleting or updating data then it should be safe to use 2nd approach.
Regards, Nitan Cell: 510 449 9629 > On Aug 13, 2020, at 11:48 AM, Pushpendra Rajpoot > <pushpendra.nh.rajp...@gmail.com> wrote: > > > Hi, > > I have a cluster of 2 DC, each DC has 5 nodes in production. This cluster is > based on active-passive model i.e. application is writing data on one DC > (Active) & it's replicated to other DC (Passive). > > My Passive DC has corrupt sstables (3 nodes out of 5 nodes) whereas there are > no corrupt sstables on any node in Active DC. > > Now, I am planning to do the following activity on affected nodes in Passive > DC: > 1. delete keyspace having corrupt sstables > 2. run 'nodetool rebuild' > > DO you see any problem in the above approach ? > > Another approach which I am considering as 'Second Option' is as given below: > 1. Stop the affected node > 2. Remove all corrupt sstables from the affected node > 3. Start the affected node > 4. Run repair on the affected node > > In my case, data is not written to Passive DC and has corrupt sstables. > > Does the 2nd approach lead to data resurrection ? > > I found in one of the threads in 'Mailing List' that data can ressurect if I > run 'nodetool repair'. Because of this reason, I considered this as a > secondary approach. > > Do you have any other approach to fix this problem OR I can go with one of > the above approaches? > > Regards, > Pushpendra
