Please see my reply in-line below. --- Lingxian Kong Senior Cloud Engineer (Catalyst Cloud) Trove PTL (OpenStack) OpenStack Cloud Provider Co-Lead (Kubernetes) On Mon, May 17, 2021 at 8:46 PM Bekir Fajkovic < bekir.fajkovic@citynetwork.eu> wrote:
Hello!
Thanks for the answer, however i am not sure that i exactly understand the approach of scheduling the backups by leveraging container running inside trove guest instance. I will try to figure out. Anyway, would this approach have a global impact, meaning only one type of schedule(s) will be applicable for all the tenants or is there still some kind of flexibility for each tenant to apply own schedules in this way? I would be very thankful for some more details, if possible :)
This is a feature needs to be discussed and well designed in the community, no actual work has been done at the moment as no one else shows interest and gets involved. I'm very happy to provide help if needed.
We have now deployed Trove in one of our regions and are going to let certain customers test the functionality and features. We currently deployed mysql, mariadb and postgresql datastores with latest production-ready datastore versions for mysql and mariadb (5.7.34 and 10.4.18) and for postgresql it is version 12.4.
That's nice!
As You might understand, we have many questions unanswered, and if there is anyone capable and willing to answer some of them we would be very thankful:
- What is next in pipe in terms of production-ready datastore types and datastore versions? - Clustering - according to the official documentation pages it is still experimental feature. When can we expect this to be supported and for what datastore types? - According to some info we received earlier, PostgreSQL 12.4 is only partially supported - what part of functionality is not fully supported here - replication or something else?
User and database management APIs are not supported as different datastores have totally different management models, it was decided to not implement such APIs in the future but let the trove users (db admin) manage users and dbs by themselves. - Creation of users and databases through OpenStack/Trove API is only
supported with mysql datastore type. When can we expect the same level of functionality for at least the other two datastore types?
See above.
- MongoDB in particular, when can this datastore type be expected to be supported for deployment? - In the case of database instance failure (for example failure due to the failure of the Compute node hosting the instance), is there any built-in mechanisms in Trove trying to automatically bring up and recover the instance, that i am not aware of?
For all your other questions not answered, they are all not implemented yet given the resources the team has. If DBaaS is in your roadmap and Trove is in your radar, I appreciate if you could get involved and start making contributions.