-
Notifications
You must be signed in to change notification settings - Fork 331
Open
Labels
area/toolingAffects the dev or user toolchain (e.g. tests, ci, build tools)Affects the dev or user toolchain (e.g. tests, ci, build tools)kind/proposalSomething fundamentally needs to changeSomething fundamentally needs to change
Description
Problem Statement
Right now you are tagging releases with latest
and vMAJOR.MINOR.PATCH
.
Could you please also tag them with vMAJOR
?
This will give us a bit more peace of mind for selfhosted environments, where we can enable auto-update for images tagged (currently) with v1
, so that when you eventually release a breaking change, the v1
tag will not pull a breaking change automatically, but will require a manual intervention.
An example of this can be seen e.g. in traefik images, whose tags include (among others) vMAJOR.MINOR.PATCH
, vMAJOR.MINOR
and vMAJOR
.
Solution Brainstorm
No response
Metadata
Metadata
Assignees
Labels
area/toolingAffects the dev or user toolchain (e.g. tests, ci, build tools)Affects the dev or user toolchain (e.g. tests, ci, build tools)kind/proposalSomething fundamentally needs to changeSomething fundamentally needs to change
Type
Projects
Milestone
Relationships
Development
Select code repository
Activity
vroldanbet commentedon Apr 23, 2025
This is doable, but I'm curious about the auto-update nature of this setup. Are you using
spicedb-operator
or running your own SpiceDB deployment? If so, how do you manage migrations, specially multi-stage migrations?maximbaz commentedon Apr 23, 2025
I'm at the stage where I don't mind a tiny downtime once in a while, so I'm using a simple docker-compose setup (very similar to what you have in your examples), with watchtower monitoring the specified tags, periodically pulling newer images and restarting containers when there is anything new. If I use
latest
, then watchtower will pulllatest
all the time, while if I could usev1
then watchtower will never automatically pull the next major version.I'm using this approach for a bunch of other projects, but I'm very early on my spicedb journey, so I don't know yet if this is an undesirable approach.
vroldanbet commentedon Apr 23, 2025
@maximbaz if this is for development, and you are using a setup that calls
spicedb migrate head
, I see no issue with this. This would become problematic for multi-phase migrations. If running in a Kube cluster and starting to look into productionizing, we recommend using thespicedb-operator
, which has the update graph baked into it and knows how to move the cluster through the updates.