Edge clusters: shared-library extraction + spec + implementation - #1225

Draft

schmidt-scaled wants to merge 31 commits into

mainfrom

Commits on Aug 6, 2026

Extract shared infrastructure into simplyblock_lib (edge-clusters step 1)

michixs
and
claude
committed
b77d9a5

Commits on Aug 7, 2026

Edge clusters: specification + implementation (spec §1-§8)

michixs
and
claude
committed
b41aa10
Edge clusters v2: fail-over/fail-back, crypto volumes, device lifecycle + e2e suite

michixs
and
claude
committed
91371f9
Edge clusters v3: adopt the product's primary/secondary lvstore processing

michixs
and
claude
committed
bf3b730

Commits on Aug 10, 2026

Merge origin/main into edge-clusters (pick up integration-test + tenacity retry fixes)
michixs
committed
4b182bb
edge: use tenacity for the RPC-up and raid-resync waits

michixs
and
claude
committed
b356fdc
e2e: make the edge campaign actually runnable (orchestrator, central leg, tier isolation)

michixs
and
claude
committed
ea72ee8
e2e: fix node bootstrap (sgdisk pkg name, script-mode imports) + repair tool

michixs
and
claude
committed
6f5afb0
e2e: correct the central CP bootstrap (real repo, env contract, sbctl install)

michixs
and
claude
committed
32d4b67
e2e: adapt the fleet to bootstrap-cluster.sh's SSH assumptions

michixs
and
claude
committed
5eb061d
e2e: install the central control plane the kubernetes-native way (helm + CRs)

michixs
and
claude
committed
be484f9
e2e: never commit the edge campaign's state file or run artifacts

michixs
and
claude
committed
244dc13

Commits on Aug 11, 2026

e2e: size the root disk and pin the subnet AZ; helm needs KUBECONFIG

michixs
and
claude
committed
1a27372
e2e: deploy the BRANCH build, not the released image

michixs
and
claude
committed
2b19a00
edge: make a failed node add diagnosable and retryable

michixs
and
claude
committed
d39506a

Commits on Aug 12, 2026

edge: provision the job ServiceAccount on the edge cluster

michixs
and
claude
committed
95689e7
edge: single admission check for node adds (API + ops)

michixs
and
claude
committed
e9fda2b
edge: disable the cpu-topology job by default

michixs
and
claude
committed
8deb8ef

Commits on Aug 13, 2026

edge_e2e: static CPU policy as k3s provisioning prerequisite

michixs
and
claude
committed
b168b59
edge_e2e: verify recorded edge cluster before reusing it

michixs
and
claude
committed
f982c65
edge: hugepages prerequisite + realistic RPC wait

michixs
and
claude
committed
cbf6d39
edge_e2e: run the released-image bootstrap phase exactly once

michixs
and
claude
committed
ece6dd8
edge: real SPDK/proxy image defaults + k8s error bodies

michixs
and
claude
committed
5875502
edge: refuse k8s access for an edge cluster without an endpoint

michixs
and
claude
committed
a2cee6e
edge: adopt the real SPDK pod launch contract

michixs
and
claude
committed
186a0f8
edge: stable identity across add retries + pod-owned hugepages + rpc retry

michixs
and
claude
committed
d27881d
edge: hugepages resource request is non-optional (cgroup gate)

michixs
and
claude
committed
83eccd1
edge: pre-init pool sizing for edge-scale memory

michixs
and
claude
committed
f04658b
edge_e2e: idempotent node adds on campaign re-run

michixs
and
claude
committed
441642a
core: hyperscale node monitor must not verdict edge clusters

michixs
and
claude
committed
42ff359

Commits on Aug 14, 2026

edge_e2e: thread-based test timeouts (Windows driver)

michixs
and
claude
committed
cc726c2