Skip to content

ci(e2e): keep every e2e leg inside the runner's disk - #18

Closed
geiresx wants to merge 5 commits into
mainfrom
pgp/e2e-disk
Closed

geiresx wants to merge 5 commits into
mainfrom
pgp/e2e-disk

Conversation

@geiresx

@geiresx geiresx commented Nov 5, 2026

Copy link
Copy Markdown
Owner

Docker's containerd store on ubuntu-26.04 keeps each layer twice, and every
leg pre-pulled all 56 images; legs on 72 GB runners ran out of disk in the
pull. Legs now switch Docker to overlay2 first, the five largest images are
pulled at their first use, and a leg with too little free disk fails before
its first case.
check-images.sh reads name:tag literals as image references that bypass
versions.env.
The 33 GB of the image table was containerd's compressed size; a probe run
measured the same three images at 24.8 GB on containerd and 17.6 GB on
overlay2.
…e at a time

7 of the 46 archives saved from containerd's store did not load into
overlay2, and an exact key hit never saves again: new keys. A 72 GB runner
reached 7 GB free with two lazy images left on it (run 41960912641,
shard 2): each now goes when its case ends.
@geiresx geiresx closed this Nov 5, 2026
@geiresx
geiresx deleted the pgp/e2e-disk branch November 5, 2026 08:39
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants