Pre-pull cache-from images

[dgageot]

Fixes #936

Signed-off-by: David Gageot [email protected]

[codecov-io]

Codecov Report

Merging #1495 into master will increase coverage by 0.05%.
The diff coverage is 56%.

@@            Coverage Diff             @@
##           master    #1495      +/-   ##
==========================================
+ Coverage   44.87%   44.93%   +0.05%     
==========================================
  Files         115      115              
  Lines        4771     4796      +25     
==========================================
+ Hits         2141     2155      +14     
- Misses       2409     2420      +11     
  Partials      221      221

Impacted Files	Coverage Δ
pkg/skaffold/docker/image.go	52.55% <0%> (-4.59%)	⬇️
pkg/skaffold/build/local/docker.go	54.83% <100%> (+37.19%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 51bcca6...88f1247. Read the comment docs.

[tjerkw]

Really nice! We build our own cacheFrom pulling code around skaffold in our ci-pipeline to ensure its fast. This fixes it. Will it also pull new versions even if it has a local image of the specified tag?

[dgageot]

Will it also pull new versions even if it has a local image of the specified tag?

Do you think it should? It currently doesn't but both behaviour would make sense to me. I guess always pulling is more predictable.

[tjerkw]

Will it also pull new versions even if it has a local image of the specified tag?

Do you think it should? It currently doesn't but both behaviour would make sense to me. I guess always pulling is more predictable.

It depends on the context, for us, in our CI environment, we always need to pull, because that local images might be stale.

However for local development, you don't want to pull everytime.

So maybe it could be a flag?
Or, if we need to choose: Dont pull if its locally available, then we can still pre-pull before calling skaffold (thats what we currently doing).

If you always pull, you have to make sure you dont pull the same cacheFrom image twice.
What we do in our scripts, is to gather all cacheFrom images in a Set, so we dont have dups, and then pull them.

[dgageot]

@tjerkw let's merge it as it is and I'll create a ticket to configure 'pull always'