I think we should not be deprecating o3fs. It is ok to make 'ofs' default. o3fs couldn't be a default anyway because it needs a bucket created. However, let's not rush to delete o3fs. We can have both ofs and o3fs around, they have their own respective advantages. A few releases down the line, we may have more data on which file system sees more adoption. For example, consider multi-tenant use cases where a customer wants isolated namespaces for two different business units. For a hybrid cloud deployments, where a user wants to migrate data between ozone buckets and s3 buckets, o3fs may turn out to be more conducive, given the file system is rooted in an ozone bucket, similar to a s3 bucket in the cloud. That said, I am +1 to following: - Offer 'ofs' as the default file system. - Update documents to reflect ofs, but keep a section for o3fs as well for folks who have tried o3fs in the past and want to continue using it. - Add ofs tests, but keep o3fs tests around so that it doesn't regress. - Evangelize 'ofs' as the primary file system offering, as it is easier to use.
On Wed, Jun 24, 2020 at 8:50 AM Wei-Chiu Chuang <weic...@cloudera.com.invalid> wrote: > Sounds good to me as long as we make sure all user docs are updated to use > o3fs as examples. > > On Tue, Jun 23, 2020 at 11:35 AM Siyao Meng <sm...@cloudera.com.invalid> > wrote: > > > Fellow Ozone developers, > > > > I am thinking of making OFS the default since it should be the way > > forward. The main reason I am sending this email is that I am about to do > > some work to deprecate o3fs and replace existing o3fs paths in the > existing > > user guides with OFS ones. Also, new tests will be added to test OFS > (e.g. > > MapReduce acceptance tests). OFS follow-up work uber jira is created > > at HDDS-3820. > > > > At least for the upcoming 0.6.0 release, both o3fs and OFS will be > usable > > on the client. o3fs won't be gone any time soon. > > > > If you have any concerns, comments and suggestions. Feel free to > discuss > > it in this thread. Thanks! > > > > Cheers, > > Siyao Meng > > >
