Skip to content

lab: use the instance store where RunsOn put it - #280

Merged
aledbf merged 3 commits into
mainfrom
lab-nvme-fix
Nov 19, 2026
Merged

aledbf merged 3 commits into
mainfrom
lab-nvme-fix

Conversation

@aledbf

@aledbf aledbf commented Nov 19, 2026

Copy link
Copy Markdown
Member

The first runs on c8id (spin-stack/ci#21) failed in lab:unpack. RunsOn builds an md RAID over the instance store and mounts it at /mnt/ephemeral, /var/lib/docker and /var/lib/containerd. The disk itself shows no mount, so findmnt -S stopped the step. Had it not, the code would have formatted a disk RunsOn was using.

  • Where the scratch goes: any mountpoint of a device on the disk is used, one a line through findmnt, preferring a mount of its own (/mnt/ephemeral) over docker's or containerd's.
  • When it formats: only a disk with no partition, holder or filesystem signature.
  • What the log shows: the layout is printed (lsblk). awk reads every line, so a closed pipe cannot fail the step under pipefail.

Run 44555369589: TMPDIR: /mnt/ephemeral/spin-lab-tmp on /dev/md0. On v20261119.01 the restore reaches the first login at 174/185 ms from local NVMe (154/161 cached), against ~840 on gp3. The 154 MiB state reads in ~40 ms.

🤖 Generated with Claude Code

aledbf and others added 3 commits November 19, 2026 12:27
… one

lab:unpack failed on the first c8id run: RunsOn had already taken the
instance store through a device on it, the disk itself showed no mount,
findmnt's exit stopped the step - and had it not, the disk would have
been formatted under RunsOn. Any mountpoint on the disk or what sits on
it is used; only a disk with no partition, holder or signature is
formatted. The layout is printed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
lsblk -r joined md0's three mountpoints with \x0a, and TMPDIR became
that string; findmnt gives one a line. awk reads to the end, so a
closed pipe cannot fail the step under pipefail.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The bench is a c8id with its scratch on local NVMe: the cold first login
333/342, the restore 174/185, one idle workspace ~125 MiB of the host.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@aledbf
aledbf merged commit cb3cc1d into main Nov 19, 2026
1 check passed
@aledbf
aledbf deleted the lab-nvme-fix branch November 19, 2026 15:52
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.

1 participant