hi, is it possible to trigger a run of an old vers...
# prefect-community
v
hi, is it possible to trigger a run of an old version of a flow? this would be useful for us in an ml ops scenario where we need full reproducibility
this is assuming the flow has been registered with prefect server, and it's persisted version is still available to be de-serialized and re-run
j
Hi @Verun Rahimtoola I don’t believe archived flows are able to be unarchived
v
hmm so basically once a new version of a flow is registered, there's no way to re-run a previous version?
prefect run flow
seems to take a
--version
argument, i just noticed
j
Ah I believe that version option may be legacy and will no longer function in that way, thanks for noticing that. It will only function properly when you call it with the most recent flow run version. Unarchiving older flows so they can be run is something that is on our radar to introduce but it is not implemented at this time
v
ok, got it. thanks for clarifying @josh
would it be possible to point us in the direction of how this might be accomplished if we absolutely needed this functionality?
perhaps we could submit a patch or something?
j
Server does have an unarchive_flow function that appear to be old https://github.com/PrefectHQ/server/blob/69e0414edc9ae738326eb2e505007b34e5c0c575/src/prefect_server/api/flows.py#L365 however it is not exposed via the GraphQL API
v
i see, thanks
may we assume that the schema of the graphql data will remain more or less stable? in case we need to run some of our own queries against it, for example
i know that the prefect cli underneath uses graphql but there are some scenarios in which we'd like to be able to run some of our own queries if possible, thanks