Comparison
Image pull policyvsImage tag
Image pull policy
you pushed a fix under the same tag, restarted the workload, and the node kept running the copy it already had.
The rule for whether a node fetches an image it already has cached. Always pulling costs registry traffic and start-up latency; never pulling makes start-ups fast and makes mutable tags actively dangerous. It is the mechanism behind the classic confusion where a rebuild appears to have no effect on some nodes and to work on others, depending on what each had cached.
Full entry →Image tag
two nodes ran what the manifest called the same version and behaved differently, because someone had pushed over the tag.
A human-readable, mutable pointer to an image — `v2.3.1`, `main`, `latest`. Because it can be repointed at any time, a tag is a name rather than an identity: pulling it twice can legitimately give you two different images. Almost every 'but it worked in staging' mystery in a container platform is a tag that moved between the two deployments.
Full entry →