dc0-edge-egress-outage-20260802.txt
====================================
The vr1-dc0 OPNsense edge has been down since the 2026-08-01 05:48 containment-VM
reboot. Found 2026-08-02 while running Step 3.5, by `dc-mirror.sh check dc0`.
Read-only throughout. Nothing was restarted by this capture.
HOW IT SURFACED -- a gate doing exactly what it was fixed to do
dc-mirror.sh check dc0 -> FAIL
"MISS last-sync FAILED: FAIL 2026-08-02T00:42:36Z ubuntu=2 uca=2
-- the mirror content is NOT attested current"
This is the 2026-07-27 fix earning its keep: before it, `check` asserted that
last-sync.status EXISTED and printed it behind an unconditional OK, so a FAIL
status passed. Content-assertion caught a 19-hour outage.
MEASURED
sync journal "rsync error: error in socket IO (code 10)"
"Download of dists/jammy-updates/caracal/Release failed:
500 Can't connect to ubuntu-cloud.archive.canonical.com"
dc0-mirror-sync.service: Failed with result 'exit-code'
rack -> archive.ubuntu.com 000
rack -> streams.canonical.com/juju/tools/ 000
rack -> api.snapcraft.io VIA the snap proxy 000
rack -> 1.1.1.1 100% packet loss
rack -> 10.12.4.1 (its OWN default gateway) 100% packet loss
ip neigh dev virbr5: 10.12.4.1 FAILED
10.12.4.5 lladdr ... STALE (juju controller, fine)
10.12.4.6 lladdr ... STALE (MAAS region, fine)
virsh: vr1-dc0-opnsense RUNNING, vnet0 on vr1-dc0-provider-public,
vnet1 on vr1-dc0-wan -- both attached
qga: "QEMU guest agent is not connected" (the G13 qga retrofit was
office1-opnsense ONLY; the DC edges never got it)
qemu log: last write 2026-08-01 05:48 -- the reboot
last mirror sync SUCCESS: 2026-08-01 00:04:19, i.e. BEFORE that reboot
SO: the edge VM is running with its NICs attached and is not answering at L2 on
its LAN leg, while two neighbours on the same segment answer normally.
WHY IT BLOCKS STAGE 5 STEP 4
apt would be fine -- the mirror still serves jammy, jammy-updates,
jammy-backports and jammy-security at 200 from its Aug-1 content.
The deploy ALSO needs streams.canonical.com for juju agent binaries and the
snap proxy's upstream for snaps. Both are dead. That is exactly the failure
class the 2026-07-31 deploy hit (four artifact defects, each masked by the one
before it).
>>> TWO GAPS THIS EXPOSES, both worth more than the incident <<<
1. NO GATE ANYWHERE TESTS DC EGRESS. `dc-rack-net.sh check dc0` PASSES with
egress dead -- it asserts bridge legs and unit states, not reachability.
preflight has no egress gate either. A 19-hour outage on the deploy's
critical path surfaced only because a nightly timer happened to fail.
2. A POST-REBOOT VERIFICATION THAT LOOKED THOROUGH AND WAS NOT. This session
reported after the 05:48 reboot that "everything reboot-persistent recovered
unaided", citing the rack units, the region API 200, the mirror answering
200, the snap proxy LISTENING and the juju controller connected. Every one
of those was true. None of them tested egress, which was the thing broken.
A listener is not a path; an HTTP 200 from a LOCAL service says nothing
about upstream. The check set was incomplete and the summary was more
confident than its evidence.
================================================================================
ROOT CAUSE FOUND -- AND IT IS NOT THE REBOOT (appended after the console read)
================================================================================
The dc0 edge is sitting at a SINGLE-USER PROMPT with missing shared libraries:
ld-elf.so.1: Shared object "libcrypto.so.17" not found, required by "php"
ld-elf.so.1: Shared object "libpython3.13.so.1.0" not found, required by "python3"
>>> Error in early script '10-configd'
>>> Error in early script '15-templates'
>>> Error in early script '90-carp'
Launching the init system...ld-elf.so.1: ... libcrypto.so.17 not found
Enter full pathname of shell or RETURN for /bin/sh:
That is a PARTIALLY-APPLIED OPNsense UPDATE. The reboot did not break the edge --
it EXPOSED it. The running system had survived on already-mapped libraries; the
2026-08-01 05:49 boot is where it fell over. Serial log mtime 05:49:39 confirms.
This symptom appears NOWHERE in appendix-A or the docs -- it is a new class.
SCOPE, MEASURED RATHER THAN ASSUMED -- IT IS dc0-ONLY
dc1 edge serial log: ZERO matches for libcrypto.so.17 / libpython3.13; it booted
clean to a login prompt with "WAN (vtnet1) -> v4: 172.30.3.2/24".
dc1 rack -> its edge 10.12.64.1 : 0% loss. vcloud -> 172.30.3.2 : 0% loss.
vcloud -> 172.30.2.2 (dc0 edge WAN) : 100% loss -- consistent with no config loaded.
THE UPSTREAM CHAIN IS HEALTHY -- the break is at the dc0 edge, nowhere else
vcloud -> 1.1.1.1 0% loss, archive.ubuntu.com 200
voffice1 -> 1.1.1.1 0% loss, archive.ubuntu.com 200, office1 edge 10.10.0.1 0% loss
both uplink nets ACTIVE, forward mode='nat', gateways 172.30.2.1 / 172.30.3.1
both pingable from vcloud
both racks hold a default route via their own edge
RECOVERY MATERIAL AVAILABLE
The pristine base EXISTS in the same pool as the broken disk:
vr1-dc0-inner-pool -> opnsense-26.7-nano.qcow2
vr1-dc0-opnsense-disk.qcow2 (the broken one)
The broken disk has an EMPTY <backingStore/> -- it is a standalone qcow2, NOT an
overlay, so there is no copy-on-write base to roll back to. A rebuild would mean
a fresh disk from the nano base plus the D-112(c) console bootstrap and D-113
API-key mint, then re-addressing (WAN 172.30.2.2/24, LAN 10.12.4.1/22).
SEPARATE THREAD, NOT CONFLATED WITH THIS ONE: dc1's rack also cannot reach 1.1.1.1
even though its edge is healthy on BOTH legs and it holds a default route via it.
That is an edge forwarding/NAT question at dc1, not the dc0 outage, and dc1 is not
on the deploy path. Logged so it is not lost; NOT diagnosed here.
================================================================================
>>> CORRECTION: THE "PARTIALLY-APPLIED UPDATE" DIAGNOSIS ABOVE IS WRONG. <<<
The real cause is UFS damage from an unclean power cut. Measured, not reasoned.
================================================================================
The section above concluded "a PARTIALLY-APPLIED OPNsense UPDATE". That was an
inference from the SHAPE of the symptom (missing libs => something replaced them)
and it is REFUTED by the log:
update / pkg-static / opnsense-update / firmware activity in the whole log: 0 lines
What the log actually shows, immediately BEFORE the first missing-library error:
UNREF FILE I=1348940 OWNER=root MODE=100644 MTIME=Jul 6 19:48 2026
RECONNECT? yes ... x2533
UNEXPECTED SOFT UPDATE INCONSISTENCY ... x785
SALVAGE? yes ... x257
** Phase 5 - Check Cyl groups
FREE BLK COUNT(S) WRONG IN SUPERBLK / BLK(S) MISSING IN BIT MAPS / SUMMARY INFORMATION BAD
***** FILE SYSTEM MARKED CLEAN *****
***** FILE SYSTEM WAS MODIFIED *****
ld-elf.so.1: Shared object "libcrypto.so.17" not found, required by "php"
fsck salvaged a badly damaged UFS root and the shared libraries did not survive it.
`libcrypto.so.17` and `libpython3.13.so.1.0` were fsck casualties.
THE CAUSAL CHAIN, end to end
1. The 2026-08-01 memory resize (416 -> 480 GiB) was an IN-PLACE tofu update, and
the repo already records that an in-place update BOUNCES the guest -- confirmed
on dc1 as the canary that day (domain id 9 -> 12, uptime 0 min).
2. A bounce of the CONTAINMENT VM is a hard power cut to every INNER VM. Nothing
shuts the inner guests down first.
3. Under D-127 the DC edges are the only inner VMs with autostart=true ("comes up
with its containment VM"). The nine role nodes were powered off; the MAAS region
and juju controller VMs are autostart=false and were started BY HAND afterwards,
cleanly. So the edges were the ONLY inner guests hard-cut and auto-restarted.
4. Both edges restarted onto damaged UFS. fsck salvaged both.
CONFIRMED BY THE dc1 CONTROL -- same event, same class, different severity:
dc0 edge dc1 edge
UNREF FILE 2533 76
SOFT UPDATE incons. 785 181
SALVAGE? yes 257 46
FILE SYSTEM MODIFIED 1 1
ld-elf errors present 0
qemu restart 05:48:31 05:45:30
dc0 lost its shared libraries -> cannot reach multi-user -> no LAN service at all.
dc1 kept its libraries -> boots to a login prompt -> BUT its damage is real too,
which is the likely explanation for the "healthy edge that does not forward"
logged earlier as a SEPARATE thread. IT IS NOT SEPARATE: same incident, lesser
damage. That earlier note should be read as superseded by this one.
WHY A "RESTART" DID THIS, stated plainly: it was not a restart of the edge. It was a
POWER CUT. The edge was never asked to shut down; its parent VM was replaced under it
while it had unflushed metadata.
THE PROCEDURE GAP, which is the durable finding:
The resize was executed with a plan asserted on CONTENT (1 in-place, 0 create/
destroy/replace, 0 MAC changes) and a capacity gate -- all correct, and none of it
covered the INNER guests. "In-place" describes the tofu resource, not the blast
radius: for a containment VM it means every VM inside it is hard-cut. Any future
in-place change to a containment VM must SHUT DOWN its inner guests first, or accept
filesystem damage to every one of them that is running.
