feat(tenants): support for tenant storage migration #1747
Conversation
Signed-off-by: Timo Glastra <timo@animo.id>
Signed-off-by: Timo Glastra <timo@animo.id>
Signed-off-by: Timo Glastra <timo@animo.id>
| * When this method is called we can be sure that we are in the mutex runExclusive lock and thus other sessions | ||
| * will not be able to open a session for this tenant until we're done. | ||
| * | ||
| * NOTE: We don't support multi-instance locking for now. That means you can only have a single instance open and |
There was a problem hiding this comment.
Interesting note. I think this applies also for single-tenancy cases where we have multiple instances. It would be another reason of disabling automatic updating in server side deployments.
There was a problem hiding this comment.
Yes agreed. It is something we can fix in a future version but you'd need something like redis to create the lock across all instances. Or we can use the askar record locking (forUpdate=true) but i'm not really sure yet how that works.
But you generally want all sessions closed when updating so if you have multiple instances in single tenancy it's similar to having one instance during the update.
Definitely room for improvement here.. Will try to tinker with it more
| @@ -9,6 +9,16 @@ import type { | |||
| } from '../types' | |||
| import type { Buffer } from '../utils/buffer' | |||
|
|
|||
| // Split up into WalletManager and Wallet instance | |||
There was a problem hiding this comment.
Ah damn. No i thought i wanted to do that now, but it's a biggie and we should move this to 0.6
Signed-off-by: Timo Glastra <timo@animo.id>
Signed-off-by: Timo Glastra <timo@animo.id>
Signed-off-by: Timo Glastra <timo@animo.id>
Signed-off-by: Timo Glastra <timo@animo.id>
This PR adds support for storage migration on in the
@credo-ts/tenantspackage. It adds checks to ensure an outdated tenants record is not used, and it also adds methods to update tenant storage. If autoUpdateStorageOnStartup is enabled it will automatically update a tenant when it is first opened (not on agent startup, but on tenant startup) to not have to update all tenants on startup.You an also query all outdated tenants, as well as update each tenant manually. From 0.6 onwards I'd like to make all breaking chagnes to storage last 2 versions, so the framework code for e.g. 0.6 can handle storage version for 0.5 and 0.6. This means you can progressively update storage and even use the tenants still while you update the storage. As now you have to go offline for a bit (remove all 0.4 agents, then spin up one 0.5 agent and update the storage), while with the other approach you can keep the 0.4 running, add in one 0.5 agent, and then scale down the 0.4 agents.