zot icon indicating copy to clipboard operation
zot copied to clipboard

[Feat]: redis backend for dedup cache

Open elee1766 opened this issue 2 years ago • 4 comments

Is your feature request related to a problem? Please describe.

i would like to use something that is not dynamodb as a distributed dedup cache.

Describe the solution you'd like

redis seems like a good candidate for this. there was already a dependency for redis in the package, so i felt it was a good choice.

Describe alternatives you've considered

No response

Additional context

No response

elee1766 avatar Nov 03 '23 08:11 elee1766

@elee1766 is this for a cloud deployment use case?

rchincha avatar Nov 03 '23 16:11 rchincha

yep!

I plan to combine this with the s3 storage

I am also currently working on the ability to use s3 storage with the sync backend.

my use case is to use zot as a proxy to balance between two identical container registries. I would like the cached images to stay existent between runs without configuring persistent storage. currently there's like a long delay on first pull which I would like to avoid.

ideally I am trying to host a distributed zot cluster using a PAAS platform, like heroku, fly, digitalocean app engine, etc. they all support redis but not dynamo since that's aws only

elee1766 avatar Nov 03 '23 20:11 elee1766

Perhaps this should be changed to Valkey or similar, since Redis isn't considered open-source anymore

AndersBennedsgaard avatar Aug 05 '24 14:08 AndersBennedsgaard

As long as it uses the redis protocol, you can install whatever redis compatible server you want (Valkey, Redis, DragonflyDB, …).

Jean-Daniel avatar Oct 01 '24 08:10 Jean-Daniel

imho redis is more adoptable for bare metal setups comparing to dynamodb, especially when scylladb is not supported (https://github.com/project-zot/zot/issues/2908)

shcherbak avatar Jan 24 '25 18:01 shcherbak