The bullet points for 500 tenants on 2.5:
- Automate tenant creation across all nodes
- Active-active read/write. Let Raven handle syncing behind the scenes; it's good at this
- More RAM is better than less RAM
- Understand how your hosting infra affects Raven. We had to write custom failover code because of all of the virtual layers on Azure (we host on Azure VMs) (https://gist.github.com/hyrmn/7262c3d7450793550e48)
- Expect replication to randomly not be set up even if it tells you it succeeded. Have logging for this (we have Raven logging to Seq)
- Pre-stage indexes. Use an index naming scheme to support this (eg,
PeopleSearch
changes get deployed asPeopleSearch2
). This means you should never reference an index by string. RavenDB 3.x with side-by-side indexes address this I think - Running is fine but server restarts are slow and messy. Block client traffic to the node and let clients failover to other node(s)
- Multi-tenant on Raven is incredibly straightforward; use it to isolate at appropri