============================================================================== AS-EXECUTED CAPTURE -- Stage 5 dc0 clean redeploy (ovn-central cert fix) Session 2026-08-04. Branch dc-dc-stage5-preconditions. ============================================================================== WHY THIS FILE: run-logged.sh opens an interactive script(1) subshell that only logs a live terminal; a Claude-driven session executes mutations as separate tool-call processes that such a subshell cannot capture. Per operator ruling 2026-08-04 ("Option 3, capture as you go"), each gated mutation's exact command and output is appended here instead -- the tool-call-friendly substitute that keeps the irreversible-step record (vault init, destroy-model) EXISTING and FINDABLE (docs/as-executed-log-convention.md purpose). CONVENTION per entry: [UTC] :: -> . Read-only measurements are NOT logged here (they are in the transcript); only mutations are. Status claims live in docs/CURRENT-STATE.md ONLY. Known coverage gap declared: read-only census/measurements this session live in the session transcript, not in this file. This file covers MUTATIONS only. ------------------------------------------------------------------------------ ## M-0 F3 redeploy prerequisite -- stage dc-node-etchosts.sh to the dc0 rack [2026-08-04] vcloud -> vr1-dc0-rack :: scp -p scripts/dc-node-etchosts.sh \ vr1-dc0-rack:/home/jessea123/repo-stage/scripts/dc-node-etchosts.sh -> rc=0 VERIFY (vr1-dc0-rack, sha256 first 16): dc-node-etchosts.sh staged=0f79de90c19e9f5a repo=0f79de90c19e9f5a MATCH mode=664 lib-net.sh (dep) staged=3e465ae7de3bf92d repo=3e465ae7de3bf92d MATCH RESULT: F3 CLOSED -- redeploy prerequisite satisfied. REVERT: rm ~/repo-stage/scripts/dc-node-etchosts.sh on the rack (restores prior staged set; the script is a redeploy input, not live cloud state). ## M.2 Path M teardown of the vr1-dc0 model [MUTATION, GA-R5 operator-ruled 2026-08-04] Operator ruling (verbatim selection): "Proceed with Path M.2 teardown now". Pre-teardown MAAS baseline (profile dc0, rack hot-kid): 10 machines, all Deployed, owner juju-vr1-dc0. juju storage empty (no storage flag). juju offers empty. [2026-08-04 ~18:2x] vr1-dc0-rack :: juju destroy-model vr1-dc0 --no-prompt (WITHOUT --force -- agents alive; --force orphaned the model + downed the controller on 08-03. WITHOUT --no-wait -- reserved for the agents-stopped stall.) DRAIN (progressive, hooks executing -- the proven path, NOT the 07-31 flat stall): 36 -> 35 -> 34 -> 30 -> 28 -> 27 -> 25 -> 24 -> 23 -> 22 -> 21 machines ... (in progress) juju models: vr1-dc0 state=destroying. DRAIN OUTCOME: graceful drain reached the tail then WEDGED flat at 15 machines (1 control host + 14 LXD containers) for ~12 min. Retry worker ALIVE (attempt 25 -> 28 -> 37 climbing) but count would not drop: 4 units stuck in workload error with failed teardown hooks their idle agents could not complete -- magnum/0 (identity-service-relation-departed), octavia/0 (certificates-relation-changed), ovn-central/0 (ovsdb-cms-relation-broken), placement/0 (identity-service-relation-departed). This is M.2's documented agents-can't-run-stop-hooks stall. Agents were EXECUTING (not stopped), so per M.2 --force was NOT auto-warranted; presented as a gated decision. INSTRUMENT NOTE (owned): first watcher counted juju status --format=json .machines and read 1 (collapses to top-level host during destroy, hiding the 14 containers) -> false STALL alarm. Real total = the "found N machines" retry line. Corrected. ## M.2b Clear the stall via juju resolved [MUTATION, GA-R5 operator-ruled 2026-08-04] Operator ruling (verbatim selection): "Try `juju resolved` on the 4 error units first". [2026-08-04] vr1-dc0-rack :: juju resolved --no-retry -m vr1-dc0 magnum/0 octavia/0 ovn-central/0 placement/0 -> rc=0 (--no-retry: skip the failing teardown hook rather than re-run it into the same failure; safe in a destroy -- MAAS re-images the machines.) RESULT: unblocked -- applications 53 -> 41 within ~30s; attempts climbing (37). New error units surface as the drain proceeds (barbican/0, designate/0 next), so continuing with a bounded resolve-loop (juju resolved --no-retry --all every ~25s) until the model is GONE (M.5 positive-signal verified). REVERT: n/a -- teardown is intended. ## M.2c Finish the stall via --force --no-wait [MUTATION, GA-R5 operator-ruled 2026-08-04] Operator ruling (verbatim selection): "Escalate: --force --no-wait now". CONDITION AT ESCALATION: resolve loop took the drain 15 -> 3 machines, then the last 3 records (control-01 + 0/lxd/2 + 0/lxd/18) went to machine-state STOPPED -- agents down, no teardown hook can run (the genuine M.2 agents-stopped stall that juju resolved cannot address). Model was 'destroying' (NOT 'dead'), so re-issuing is the documented escalation, not the forbidden dead-model re-issue. [2026-08-04 ~19:29] vr1-dc0-rack :: juju destroy-model vr1-dc0 --force --no-wait --no-prompt -> "Destroying model" ... 3 -> 1 -> 0 machines ... "Model destroyed." rc=0 RESULT (M.2 overall): TEARDOWN COMPLETE, CLEAN, NO ORPHAN. ORPHAN CHECK (the 08-03 failure -- did NOT recur): - juju models: only 'controller' (available); vr1-dc0 fully gone, not life:dead. - controller model API responds (juju status -m controller OK) -- 17070 up. - models now = [('admin/controller','alive')] -- no dead-model residue; add-model will not be refused. M.5(b) CASCADE CHECK (profile dc0): 10 machines, count UNCHANGED from baseline -> NO cascade. States: Ready 9 (role nodes, owner=None) + Deployed 1 (controller VM subtle-grouse). Exactly the expected post-teardown state. REVERT: n/a -- teardown is intended; rebuild is M.6 (add-model + rebuild). ------------------------------------------------------------------------------ ## M.5(c) proxy-survival post-checks (read-only, logged for the teardown record) dc-cache-proxy check dc0: PASS (apt-cacher-ng active :3142, serves archive+UCA 200) dc-snap-proxy check dc0: PASS (:3129, store API 200, CDN tunnel allowed, allowlist enforced) => M.5 ALL GREEN. Path M teardown complete + clean. Proceeding to M.6 rebuild. ## M.6 REBUILD ## Step 3.5 -- add-model + spaces + artifact source [MUTATION, gated] [2026-08-04] vr1-dc0-rack (D-138; runbook says voffice1 but that predates D-138) :: juju add-model vr1-dc0 -> "Added 'vr1-dc0' model with credential 'vr1-dc0-cred' for user 'admin'" rc=0 GATES (read-only): (a) artifact source inherited from model-defaults: apt-http-proxy=http://10.12.8.4:3142, snap-http[s]-proxy=http://10.12.8.4:3129, apt-mirror EMPTY. Correct per D-135 amendment (b) (dc0 on the caching PROXY, not the retired mirror). apt-mirror NOT set (Step 3.5's mirror text is STALE -- runbook-fold gap, flagged). (b) juju-spaces-check.sh vr1-dc0: PASS 0 fatal -- all six spaces present, none stale. REVERT: juju destroy-model vr1-dc0 (empty model; trivial). ## QUEUED -- NOT EXECUTED THIS SESSION (operator request 2026-08-04, mid-redeploy) Q1. "Queue a Tailscale installation for DC0 and DC1." DEFERRED (hard rule 1 -- new scope is logged mid-step, not run). NEEDS A RULING BEFORE EXECUTION, not just an ops task: it intersects ruled surfaces -- - D-107: Tailscale is the operator WORKSTATION / human-GUI path, NOT the DC nodes. - D-128: durable DC reach is the base-leg oneshot, explicitly NOT tailscale-on-host. - 2026-08-02 ruling: "VPN deferred to Roosevelt" (VPN class deferred). ACTION OWED: grep the exact D-107/D-128 text + the 08-02 VPN-deferred ruling, then put the DC-Tailscale question to the operator as a GA-R5 exchange (likely a new D-number, GA-R3 A1: decides DC access at every future standup). Do NOT install until ruled. ## Step 1.2b -- metal-internal /etc/hosts cloudinit (the ovn-central cert fix) [MUTATION, gated] FIRST ATTEMPT FAILED -- delivery-mechanism defect found: juju model-config REJECTED the config: "cloudinit-userdata: runcmd not allowed, use preruncmd or postruncmd instead" (rc=1, nothing set). The 08-04 fix proved the CONCEPT (manual /etc/hosts -> cert) but its DELIVERY (cloudinit-userdata) was never live-tested; the harness graded rendered YAML (cloud-init accepts runcmd) not juju acceptance (juju forbids it). Classic can't-see-the-live-apply green harness. FIX (repo change, delivered with discipline): scripts/dc-node-etchosts.sh: render key runcmd: -> postruncmd: (repo sha 0f79de90.. -> 107fc197..); + inline comment forbidding revert. tests/dc-node-etchosts: switched to postruncmd; NEW T10 asserts a juju-accepted key + rejects bare runcmd (the missing gate). Harness 10/10; T10 MUTATION-PROVEN to fail (revert to runcmd -> 5 FAIL). Re-staged to rack (sha 107fc197 verified). RE-RUN (live delivery test that was missing): [2026-08-04] vr1-dc0-rack :: JUJU_MODEL=vr1-dc0 bash scripts/dc-node-etchosts.sh apply vr1-dc0 -> "OK: set" (juju ACCEPTED postruncmd) JUJU_MODEL=vr1-dc0 bash scripts/dc-node-etchosts.sh check vr1-dc0 -> OK, carries 10.12.12.0/22 mapping, rc=0 GATE: check exit 0. On-node /etc/hosts delivery VERIFY-AT-PROVISION after Step 1.4. REVERT (model-config): juju model-config -m vr1-dc0 --reset cloudinit-userdata. ## Step 4.2 dry-run gate (read-only, logged for the deploy record) 9 machines "0"-"8", every machine tags=openstack-vr1-dc0, (3 control/2 compute/4 storage); ceph-osd->5-8, nova-compute->3-4; no overlay error; rc=0. GATE PASS. ## Step 4.3 -- the real deploy [MUTATION, GA-R5 operator-ruled 2026-08-04] Operator ruling (verbatim selection): "Proceed with the deploy". [2026-08-04] vr1-dc0-rack :: cd ~/repo-stage && juju deploy ./bundle.yaml \ --overlay overlays/vr1-dc0-vips.yaml --overlay overlays/vr1-dc0-machines.yaml \ --overlay overlays/vr1-dc0-octavia-pki.yaml -m vr1-dc0 -> "Deploy of bundle completed." rc=0. 9 machines + ~24 containers/units added. ovn-central 0/lxd/15, 1/lxd/3, 2/lxd/3 (will provision WITH the postruncmd /etc/hosts fix). REVERT: juju destroy-model vr1-dc0 (Path M). ## Step 4.4 -- merged-config gate (read-only, run immediately) GATE PASS ovn-chassis all THREE options present (source=user): bridge-interface-mappings= 'br-ex:52:54:00:8c:2a:8c br-ex:52:54:00:50:48:88', ovn-bridge-mappings='physnet1:br-ex', prefer-chassis-as-gw=True -- juju merged the options map key-by-key correctly. openstack-origin/source: all set, none UNSET (mix of cloud:jammy-caracal + caracal, matching the prior converged deploy / same bundle sha -- observation, not a defect). First execution of this gate ran clean (no shape correction needed). ## CONVERGE INCIDENT -- mysql-innodb-cluster/1 wedged (08-03 apt recurrence) [operator-ruled] SYMPTOM: mysql/1 (container 1/lxd/2) stuck agent=allocating "waiting for machine" ~90min. cloud-init blocked since 20:24:54 on `eatmydata apt-get update` (capture=False) that infinitely retried a hanging index (jammy-backports InRelease via the proxy) and never completed -> jujud never installed (/var/lib/juju/tools empty). The 08-03 apt-wedge class; appendix-A has NO entry for it (DOC GAP owed -- fold the symptom+fix into appendix-A). REMEDY 1 (operator ruling: "Restart apt-cacher-ng proxy first") [MUTATION]: vr1-dc0-rack :: sudo systemctl restart apt-cacher-ng -> active, :3142. Proxy PROVEN healthy after (curl -x proxy jammy-backports InRelease = 200 in ~1s). But the container's apt stayed independently wedged (stale looping process); restart alone did not clear it. REMEDY 2 (operator ruling: "Remove + re-add the unit") [MUTATION]: vr1-dc0-rack :: juju remove-unit mysql-innodb-cluster/1 --force --no-prompt -> rc=0 (--force: agent never came up, no hooks to run; unit never joined the cluster). Verified: /1 gone, container 1/lxd/2 GONE. vr1-dc0-rack :: juju add-unit mysql-innodb-cluster -n 1 --to lxd:1 -> rc=0 -> mysql-innodb-cluster/3 on fresh container 1/lxd/4 (provisioning against the healthy proxy). REVERT: n/a (converge remediation; the fresh unit replaces the wedged one). ## >>> MAJOR FINDING (verify-at-provision gate CAUGHT IT): the D-052 amendment binding ## (53e046e) makes ovn-central UNREACHABLE from the controller. NEVER DEPLOYED BEFORE. <<< MEASURED: all 3 ovn-central containers have a SINGLE leg on the ISOLATED metal-internal plane (/0=10.12.12.106, /1=10.12.12.3, /2=10.12.12.1), NO default route, NO metal-admin leg. -> cannot reach the controller (10.12.8.5/10.12.4.5 :17070 UNREACHABLE) -> "download agent binaries" retries forever (attempt 547+) -> jujud installed but agent stuck allocating -> ovn-central can NEVER come up. mysql containers have TWO legs (metal-admin + metal-internal); ovn-central got metal-internal ONLY because the bundle binds EVERY ovn-central endpoint incl the DEFAULT '' to metal-internal. ROOT: commit 53e046e (2026-08-03 20:49) "apply ovn-central default -> metal-internal": '': metal-admin -> '': metal-internal. juju uses the DEFAULT binding for the machine's controller/management connection, so moving it to the isolated plane cut controller reach. This amendment was NEVER DEPLOYED: the 08-03 deploy PREDATED it (its ovn-central units were idle/up, degraded on cert only -- which REQUIRES controller reachability, impossible on a metal-internal-only leg). This fresh deploy is its first exposure; it fails. CROSS-CUT: the 08-04 /etc/hosts postruncmd fix (this session, PROVEN logic -- manual WOULD-ADD test) makes get_hostname resolve INDEPENDENTLY of the binding. So the amendment's stated purpose (single-space cert-CN resolution) is now covered by a different mechanism, and the amendment's remaining effect is purely harmful (unreachable ovn-central). => This is a RULED-BUT-NOT-BUILT amendment whose deciding reason is refuted by measurement (the D-139 shape). Needs a GA-R5 D-052 RE-AMENDMENT ruling. HOLD vault init until ruled (settle gate PROCEED is true-but-irrelevant: it never checks ovn-central). LOGGED (not fixed, hard rule 1): (a) do NOT edit bundle bindings mid-step -- present options; (b) the postruncmd mechanism is STRUCTURALLY FRAGILE -- it runs at the END of juju's runcmd, so ANY earlier blocker (apt wedge, agent-binary retry) prevents the /etc/hosts entry from ever being written. Delivery is gated behind juju's entire runcmd completing. Record in the changelog/appendix-A regardless of the binding ruling. Model state PRESERVED (converged-but-ovn-blocked) for whichever way the operator rules. ## D-052 RE-AMENDMENT APPLIED -- ovn-central re-homed to metal-admin [MUTATION, GA-R5 2026-08-05] Operator ruling (exact utterance): "Once that is done then move on to the proposed move to metal-admin". Repo half committed d6975fb (bundle '' -> metal-admin; D-052 re-amendment; reference doc). Live half: [2026-08-05] vr1-dc0-rack :: scp -p bundle.yaml -> rack ~/repo-stage (sha 213b51baea23a3f9 verified) juju remove-application ovn-central --force --no-prompt -> rc=0 (3 stuck units removed; agents never came up, so --force clean -- containers were single-legged on metal-internal, unremovable in place) DRY-RUN re-deploy: plan = add ovn-central + 4 relations (certificates, ovsdb-cms, ovsdb x2) + 3 units ONLY; no other app touched. Confirmed before applying. juju deploy ./bundle.yaml --overlay {vips,machines,octavia-pki} -m vr1-dc0 -> "Deploy of bundle completed" rc=0. ovn-central/3,4,5 on fresh 0/lxd/0, 1/lxd/0, 2/lxd/0 with '' = metal-admin. GATE (pending verify): new containers must have a metal-admin leg (10.12.8.x) -> controller reachable -> agent starts (NOT stuck allocating). Then /etc/hosts postruncmd at cloud-init end. REVERT: n/a (this IS the fix; the prior binding is superseded by the D-052 re-amendment). ## >>> D-052 RE-AMENDMENT VALIDATED LIVE -- BOTH ovn-central FIXES CONFIRMED. <<< (1) REACHABILITY: ovn-central/3,4,5 agents went allocating -> executing (poll 5). New containers are TWO-LEGGED: ovn-central/3 (0/lxd/19) eth0 10.12.9.14 (metal-admin) + eth1 10.12.12.129 (metal-internal). The metal-admin leg = controller reachable = agent starts. The single-leg failure is CLOSED by the '' -> metal-admin re-amendment. (2) CERT-CN DELIVERY: /etc/hosts on ovn-central/3 carries "10.12.12.129 juju-d4f45a-0-lxd-19.maas juju-d4f45a-0-lxd-19" and getent hosts reverse-RESOLVES it -> get_hostname(metal-internal addr) will return a non-empty CN -> vault will issue the server cert. This is the postruncmd DELIVERY PROOF that was impossible on 08-04 (the containers were wedged/mis-bound). BOTH fixes work end-to-end pre-cert. PRE-VAULT-INIT SETTLE RE-CONFIRMED after the re-home: phase-02-vault-preflight vr1-dc0 -> PROCEED (mysql 3/3 active+ONLINE 1 R/W, vault fresh, census 66 units 0 workload-error 0 agent-error, 40 active). Model ready for vault init. ## HANDED OFF (operator-only one-shot): phase-02 vault init. runbooks/phase-02-vault-bringup.md Step 2.1 (init/unseal/root-CA), run VERBATIM by the operator on the DC rack (D-138 juju client); guard-hook blocks the agent (DOCFIX-006/D-069). On completion: vault issues certs -> ovn-central requests its server cert with the resolved CN -> OVN NB/SB cluster forms (6641/6642 listen) -> the redeploy's purpose is achieved. ## (delivery proof) -> phase-02 Step 2.1 vault init (OPERATOR-ONLY one-shot) -> ovn cert ## issues -> converge. (below appended as each further mutation executes) ## >>> VAULT INIT COMPLETE (operator-run one-shot) + ovn-central RESOLVED end-to-end. 2026-08-05 <<< Operator ran phase-02-vault-bringup.md Steps 2.1-2.3 VERBATIM on the dc0 rack (D-138), model -m vr1-dc0 (VR0 -m openstack head-substituted -- fold item F13 / D-142 R1). The agent did NOT run these (guard-hook, secret custody); recorded here from the operator's pasted transcript with ALL secret material excluded. 2.1 init: PROBE_OK (pre-init writability probe, D-142 safety add, run manually); pre-check Initialized=false; init one-shot with `2>&1 | tee ~/vault-init/init.txt` -> GATE: 5 unseal keys + TOKEN_OK. Off-host save handled by the operator by hand (R2 transport gap surfaced verbally at hand-off; still the D-142 open item). 2.2 unseal: 1/3 -> 2/3 -> 3/3 -> Sealed=false, Initialized=true, Storage=mysql, HA=false; Cluster vault-cluster-12969e82. 2.3 authorize + root CA: authorize-charm op1/task2 completed; generate-root-ca op3/task4 returned the root CA PEM. vault/0 -> active/idle "Unit is ready (active: true, mlock: disabled)"; vault-mysql-router/0 active. EXIT GATE MET. ROOT CA (PUBLIC -- no secret custody; as-built metadata only): subject "Vault Root Certificate Authority (charm-pki-local)"; validity 2026-08-05 02:05:57 -> 2036-08-02 01:06:27 UTC (~10y anchor). PEM body NOT stored (public but unnecessary). SECURITY HYGIENE NOTE: the child token printed by `vault token create -field=token` appeared in plaintext in the operator's paste (init keys + root token WERE redacted). It was ttl=10m, unset after use, and long expired by paste time -> dead/benign, no rotation. Lesson: redact the token-create output too. NOT stored/echoed/committed anywhere. ## >>> ovn-central CERT CASCADE -- the redeploy's PURPOSE -- ACHIEVED (measured, read-only). <<< Post-init `juju status -m vr1-dc0`: ovn-central/3,4,5 ALL active/idle "Unit is ready"; /4 leader (ovnnb_db, ovnsb_db); /5 northd active. The OVN NB/SB cluster FORMED. This closes the multi-session ovn-central saga: the CN-from-reverse-DNS-on-isolated-plane root cause (08-04) is defeated by BOTH fixes landed this cycle -- the /etc/hosts postruncmd (CN delivery, dc-node-etchosts.sh Step 1.2b) AND the D-052 re-amendment '' -> metal-admin (reachability). Census moved 40 -> 47 active, 0 workload-error, 0 agent-error. ## REMAINING (separate item, NOT vault/ovn, logged not fixed -- hard rule 1): ceph-mon/2 + ceph-radosgw/0 still `allocating`/"waiting for machine" (container provisioning, likely the 08-03 apt-cacher-ng/cloud-init wedge class). Cascade: ceph-mon can't reach 3-peer quorum -> ceph-osd x4 "Incomplete relation: monitor" -> glance + nova-compute x2 "Incomplete storage-backend". To be triaged AFTER the bookend. Expected-waiting (not stuck): designate (Stage-7 DNS), octavia (end-user configure-resources), ceph-rbd-mirror (two-DC).