When in doubt, repair. Nodetool snapshot won't be perfectly consistent across all the nodes anyway since you're snapshotting on each node individually.
On Tue, Oct 29, 2019, 8:31 PM Ankit Gadhiya <ankitgadh...@gmail.com> wrote: > Thanks Paul. This is interesting. > So, anything I need to do after cp? - nodetool repair? > Also I am assuming I need to be doing this exercise on all the nodes of > the cluster - right? > Any suggestion to automate this or do it just from a single node? > > > — Ankit Gadhiya > > On Tue, Oct 29, 2019 at 11:21 PM Paul Carlucci <paul.carlu...@gmail.com> > wrote: > >> Straight up Unix cp command, make sure you're in the right directory. If >> you try to use schema.cql then you're going to have to massage it somewhat >> due to keyspace name differences and schema changes over time. You'll see >> what I mean if you've got some. >> >> It goes without saying that you're gonna want to try this in non-prod >> first. On a positive note you'll be learning some stuff they don't quite >> teach in Datastax Academy! >> >> On Tue, Oct 29, 2019, 8:39 AM Ankit Gadhiya <ankitgadh...@gmail.com> >> wrote: >> >>> Thanks Paul. >>> >>> Copy SSTable - How? Using SSTableLoader or some other mechanism. >>> >>> >>> *Thanks & Regards,* >>> *Ankit Gadhiya* >>> >>> >>> >>> On Tue, Oct 29, 2019 at 11:36 AM Paul Carlucci <paul.carlu...@gmail.com> >>> wrote: >>> >>>> Copy the schema from your source keyspace to your new target keyspace, >>>> nodetool snapshot on your source keyspace, copy the SSTable files over, do >>>> a rolling bounce, repair, enjoy. In my experience a rolling bounce is >>>> easier than a nodetool refresh. >>>> >>>> It's either that or just copy it with Spark. >>>> >>>> On Tue, Oct 29, 2019, 11:19 AM Ankit Gadhiya <ankitgadh...@gmail.com> >>>> wrote: >>>> >>>>> Thanks Alex. So How do I copy SSTables from 1.0 to 2.0? (Same >>>>> SSTableLoader or any other approach?) >>>>> Also since I've multi-node cluster - I'll have to do this on every >>>>> single node - is there any tool or better way to execute this just from a >>>>> single node? >>>>> >>>>> *Thanks & Regards,* >>>>> *Ankit Gadhiya* >>>>> >>>>> >>>>> >>>>> On Tue, Oct 29, 2019 at 11:16 AM Alex Ott <alex...@gmail.com> wrote: >>>>> >>>>>> You can create all tables in new keyspace, copy SSTables from 1.0 to >>>>>> 2.0 tables & use nodetool refresh on tables in KS 2.0 to say Cassandra >>>>>> about them. >>>>>> >>>>>> On Tue, Oct 29, 2019 at 4:10 PM Ankit Gadhiya <ankitgadh...@gmail.com> >>>>>> wrote: >>>>>> >>>>>>> Hello Folks, >>>>>>> >>>>>>> Greetings!. >>>>>>> >>>>>>> I've a requirement in my project to setup Blue-Green deployment for >>>>>>> Cassandra. E.x. Say My current active schema (application pointing to) >>>>>>> is >>>>>>> Keyspace V1.0 and for my next release I want to setup Keysapce 2.0 (with >>>>>>> some structural changes) and all testing/validation would happen on it >>>>>>> and >>>>>>> once successful , App would switch connection to keyspace 2.0 - This >>>>>>> would >>>>>>> be generic release deployment for our project. >>>>>>> >>>>>>> One of the approach we thought of would be to Create keyspace 2.0 as >>>>>>> clone from Keyspace 1.0 including data using sstableloader but this >>>>>>> would >>>>>>> be time consuming, also being a multi-node cluster (6+6 in each DC) - it >>>>>>> wouldn't be very feasible to do this manually on all the nodes for >>>>>>> multiple >>>>>>> tables part of that keyspace. Was wondering if we have any other >>>>>>> creative >>>>>>> way to suffice this requirement. >>>>>>> >>>>>>> Appreciate your time on this. >>>>>>> >>>>>>> >>>>>>> *Thanks & Regards,* >>>>>>> *Ankit Gadhiya* >>>>>>> >>>>>>> >>>>>> >>>>>> -- >>>>>> With best wishes, Alex Ott >>>>>> http://alexott.net/ >>>>>> Twitter: alexott_en (English), alexott (Russian) >>>>>> >>>>> -- > *Thanks & Regards,* > *Ankit Gadhiya* > >
