Skip to content

Commit d8a21c1

Browse files
committed
ci(Jenkinsfile): drop two leftover diagnostics from the pipeline cleanup
- ensureImage: remove the `ls -la <ctx>` after staging the kaniko build context (miss-path success noise). - Deploy: remove the /proc/net/route + resolv.conf + per-lease TCP/22 reachability dump. It was purely observational -- nothing downstream depended on it, and a dead leased host still fails loudly when the next scp fails under set -eu. Kept a short note preserving the build #59 dead-host lesson.
1 parent 0592eaa commit d8a21c1

1 file changed

Lines changed: 7 additions & 24 deletions

File tree

ci/Jenkinsfile

Lines changed: 7 additions & 24 deletions
Original file line numberDiff line numberDiff line change
@@ -702,29 +702,13 @@ spec:
702702
# ssh's UNPROTECTED PRIVATE KEY FILE check is happy.
703703
install -m 0600 /home/jenkins/agent/.ssh-stage/key /home/jenkins/agent/ssh-key
704704
SSH_KEY=/home/jenkins/agent/ssh-key
705-
# Build #59 hit "ssh: connect to host 10.46.64.24 port 22:
706-
# No route to host" -- a dead Ampere host masquerading as a
707-
# routing problem (verified out-of-band via SSA-authenticated
708-
# ping/colossus bm host coldreboot/reprovision on 2026-05-20).
709-
# Keep the diagnostic in tree so future flakes surface
710-
# cheaply: /proc/net/route + per-lease TCP/22 probes via the
711-
# bash /dev/tcp builtin (image lacks iproute2 and netcat).
712-
echo "==== Deploy connectivity diagnostics ===="
713-
echo "---- /proc/net/route ----"
714-
cat /proc/net/route || true
715-
echo "---- /etc/resolv.conf ----"
716-
cat /etc/resolv.conf || true
717-
echo "---- per-lease reachability (TCP 22, 5s timeout) ----"
718-
while IFS= read -r TARGET; do
719-
IP=$(echo "${TARGET}" | cut -d, -f3)
720-
ARCH=$(echo "${TARGET}" | cut -d, -f5)
721-
if timeout 5 bash -c "exec 3<>/dev/tcp/${IP}/22 && exec 3>&-" 2>&1; then
722-
echo "${ARCH} @ ${IP}:22 -- reachable"
723-
else
724-
echo "${ARCH} @ ${IP}:22 -- UNREACHABLE (exit $?)"
725-
fi
726-
done < leases.txt
727-
echo "==== end diagnostics ===="
705+
# A dead leased host (build #59: an Ampere box stuck at "No
706+
# route to host" after a flaky bm reboot) surfaces here as the
707+
# scp below failing under set -eu, which fails the Deploy stage
708+
# loudly. We used to print /proc/net/route + resolv.conf + a
709+
# per-lease TCP/22 reachability dump first; dropped as console
710+
# noise -- the scp error already localizes a dead host (just
711+
# after ConnectTimeout below rather than a fast 5s probe).
728712
SSH_OPTS="-i ${SSH_KEY} -o IdentitiesOnly=yes -o PasswordAuthentication=no -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no -o ConnectTimeout=300 -o ServerAliveInterval=2 -o ServerAliveCountMax=1"
729713
BZIMAGE="${WORKSPACE}/arch/x86_64/boot/bzImage"
730714
INITRD="${BUILDROOT_OUT}/images/rootfs.cpio"
@@ -1085,7 +1069,6 @@ def ensureImage(Map pins) {
10851069
set -eu
10861070
rm -rf '${ctx}'
10871071
ci/build-image.sh --stage-only '${ctx}' --tag '${tag}'
1088-
ls -la '${ctx}'
10891072
"""
10901073
}
10911074

0 commit comments

Comments
 (0)