0.6.0 to 0.6.1 Upgrade

The 0.6.1 release replaces the archived Bitnami Cassandra runtime with the official Apache Cassandra 5.0.8 image and an in-house Cassandra Helm chart. This removes the CVE-affected Bitnami runtime and its bundled components.

Fresh installs need no special handling. Existing clusters that already hold data need a migration, because three things change at once: the runtime image, the Helm chart shape, and the on-disk data layout. A normal in-place Helm upgrade does not carry an existing Bitnami-based cluster onto the new stack:

• The new Cassandra chart intentionally removes the app.kubernetes.io/name and app.kubernetes.io/instance labels from volumeClaimTemplates.metadata.labels. This is the only change to an immutable StatefulSet field. Kubernetes treats the entire volume claim template as immutable, so the StatefulSet has to be recreated.
• The Bitnami image stored data nested one directory deeper than the official image expects (/bitnami/cassandra/data/... versus /var/lib/cassandra/...), so the new image does not find the old data by default.
• The old and new node configuration must agree on the settings the existing data was written with.

Because of this, an operator with an existing data-bearing cluster chooses a migration method. This guide describes the methods and their trade-offs; none is mandated over the others.

Select the bundles and cluster

Download each version into a separate directory. Keep your deployed 0.6.0 configuration and credentials; a download does not contain your site settings. Use a new directory outside the cluster for the evidence and backups. It must have capacity for all node snapshots and use your approved backup storage.

set -euo pipefail
set +x
umask 077
helm version --short | grep -Eq '^v3\.'
helmfile version --short | grep -Eq '^v?1\.1\.[0-9]+($|[+-])'
export HELMFILE_ENV="your-existing-environment"
export CONTROL_KUBECONFIG="/absolute/path/to/control-plane-kubeconfig"
export CONTROL_CONTEXT="your-reviewed-control-plane-context"
export UPGRADE_DIR="/absolute/path/to/new-upgrade-directory"
if [ -e "$UPGRADE_DIR" ]; then
  echo "Directory already exists. Use its saved session for diagnostics; do not overwrite the baseline." >&2
  exit 1
fi
mkdir "$UPGRADE_DIR"
cd "$UPGRADE_DIR"
for version in 0.6.0 0.6.1; do
  ngc registry resource download-version \
    "nvidia/nvcf/nvcf-self-managed-stack:${version}"
  mkdir "stack-${version}"
  tar -xzf \
    "nvcf-self-managed-stack_v${version}/nvcf-self-managed-stack-${version}.tar.gz" \
    -C "stack-${version}"
done
export SOURCE_CONTROL="$UPGRADE_DIR/stack-0.6.0"
export TARGET_CONTROL="$UPGRADE_DIR/stack-0.6.1"
export BACKUP_DIR="$UPGRADE_DIR/backup"
mkdir "$BACKUP_DIR"
kcp() {
  kubectl --kubeconfig "$CONTROL_KUBECONFIG" \
    --context "$CONTROL_CONTEXT" "$@"
}
test "$(kubectl --kubeconfig "$CONTROL_KUBECONFIG" config current-context)" \
  = "$CONTROL_CONTEXT"
kcp cluster-info
CONTROL_CLUSTER_UID=$(kcp get namespace kube-system -o jsonpath='{.metadata.uid}')
test -n "$CONTROL_CLUSTER_UID"
assert_context() {
  test "$(kubectl --kubeconfig "$CONTROL_KUBECONFIG" config current-context)" \
    = "$CONTROL_CONTEXT" &&
  test "$(kcp get namespace kube-system -o jsonpath='{.metadata.uid}')" \
    = "$CONTROL_CLUSTER_UID"
}
save_session() {
  local name session
  session=$(mktemp "$UPGRADE_DIR/.session.XXXXXX")
  {
    printf '%s\n' 'set -euo pipefail' 'set +x' 'umask 077' 'unset CQL_USER CQL_PASSWORD'
    for name in HELMFILE_ENV CONTROL_KUBECONFIG CONTROL_CONTEXT CONTROL_CLUSTER_UID \
      UPGRADE_DIR SOURCE_CONTROL TARGET_CONTROL BACKUP_DIR CHECKER CQL_POD \
      SNAPSHOT_TAG TARGET_CASSANDRA_IMAGE; do
      if declare -p "$name" >/dev/null 2>&1; then declare -p "$name"; fi
    done
    for name in kcp assert_context save_session capture_members load_cql_credentials \
      cql cql_json capture_data; do
      if declare -F "$name" >/dev/null; then declare -f "$name"; fi
    done
  } > "$session"
  mv "$session" "$UPGRADE_DIR/session.sh"
}
save_session

Keep this shell open for the remaining commands. The bundle Makefile accepts a kubeconfig, but no context flag. assert_context must succeed immediately before each make install. Do not run concurrent maintenance or change this kubeconfig during the procedure. Stop on any command failure; do not skip a failed check or resume at a later mutation.

Reopen a session for diagnostics

The commands save the pinned cluster identity and available helpers in $UPGRADE_DIR/session.sh, adding the snapshot tag and target image when they are known. This private file excludes credentials. If a check closes the shell after setup, open a new Bash shell and source the saved session at its prompt:

set -euo pipefail
set +x
export UPGRADE_DIR="/absolute/path/to/the-existing-upgrade-directory"
source "$UPGRADE_DIR/session.sh"
assert_context
kcp -n cassandra-system get pods

Use this session only for read-only diagnostics until the failed check is resolved and the recovery procedure is reviewed. Do not rerun setup, recapture the original baseline, or repeat mutations to bypass the failure. If CQL setup was reached, run load_cql_credentials before CQL diagnostics; it reloads the credentials from the protected baseline files, even after the old Secret is removed. A changed context or cluster UID must be investigated, not repinned.

Prepare the 0.6.1 configuration

Run cd "$TARGET_CONTROL" before preparing the configuration or running a 0.6.1 sync command. Reconcile the site-specific settings from the 0.6.0 deployment into the 0.6.1 configuration files:

• environments/$HELMFILE_ENV.yaml
• secrets/$HELMFILE_ENV-secrets.yaml

Start from the files or templates included with 0.6.1 and carry forward the required values from 0.6.0. Do not replace the 0.6.1 files wholesale because available settings can change between releases. Preserve the existing registry, storage, endpoint, credential, and secret values unless this procedure explicitly instructs you to change them.

Confirm that both files exist before continuing:

cd "$TARGET_CONTROL"
for file in \
  "environments/${HELMFILE_ENV}.yaml" \
  "secrets/${HELMFILE_ENV}-secrets.yaml"; do
  [ -s "$file" ] || { echo "missing or empty required file: $file"; exit 1; }
done

Fresh installs

A new 0.6.1 install has no existing Cassandra data and needs no migration. Install the stack normally; Cassandra comes up on the official image with the in-house chart:

assert_context
make install HELMFILE_ENV="$HELMFILE_ENV" \
  KUBECONFIG_FILE="$CONTROL_KUBECONFIG" HELMFILE_SELECTOR=

The rest of this guide applies only when migrating an existing data-bearing cluster.

Check Cassandra before migrating

Download the read-only checker to $UPGRADE_DIR/cassandra-upgrade-check.py. It uses only the Python standard library and validates captured evidence; it does not contact or change a cluster. The checker rejects unsupported layouts and incomplete results.

export CHECKER="$UPGRADE_DIR/cassandra-upgrade-check.py"
save_session
test -s "$CHECKER"
python3 "$CHECKER" bundle "$SOURCE_CONTROL" 0.6.0
python3 "$CHECKER" bundle "$TARGET_CONTROL" 0.6.1
capture_members() {
  local directory=$1 pod
  mkdir "$directory"
  assert_context
  kcp -n cassandra-system get statefulset cassandra -o json > "$directory/statefulset.json"
  kcp -n cassandra-system get pods -o json > "$directory/pods.json"
  kcp -n cassandra-system get pvc -o json > "$directory/pvcs.json"
  for pod in cassandra-0 cassandra-1 cassandra-2; do
    kcp -n cassandra-system exec "$pod" -c cassandra -- nodetool status \
      > "$directory/$pod-ring.txt"
  done
}
save_session
capture_members "$BACKUP_DIR/before"
python3 "$CHECKER" topology "$BACKUP_DIR/before" > "$BACKUP_DIR/topology-before.json"
helm --kubeconfig "$CONTROL_KUBECONFIG" --kube-context "$CONTROL_CONTEXT" \
  -n cassandra-system list --all --filter '^cassandra$' --output json > "$BACKUP_DIR/cassandra-release.json"
jq -e 'length == 1 and .[0].name == "cassandra" and .[0].namespace == "cassandra-system"
  and .[0].status == "deployed" and .[0].chart == "helm-nvcf-cassandra-0.15.5"' \
  "$BACKUP_DIR/cassandra-release.json" >/dev/null
helm --kubeconfig "$CONTROL_KUBECONFIG" --kube-context "$CONTROL_CONTEXT" \
  -n cassandra-system get values cassandra --all -o json > "$BACKUP_DIR/cassandra-values.json"
kcp -n cassandra-system get secret cassandra -o json > "$BACKUP_DIR/cassandra-secret.json"
kcp -n cassandra-system get jobs -o json > "$BACKUP_DIR/jobs-before.json"

Success requires exactly the three expected pods, owned by the current StatefulSet, Running with every container Ready. Each node must report the same three UN (Up/Normal) ring members, matching the pod IPs. Each original data PVC must be Bound, without a deletion owner or Delete retention policy. The saved PVC/PV and Cassandra host identities are compared after adoption.

The standard 0.6.0 superuser name is in cassandra.dbUser.user in the Helm values; its live password is in Secret cassandra, key cassandra-password. If the installation uses dbUser.existingSecret, a mapped key, password files, or a customized authentication/TLS setup, stop and adapt the site procedure before continuing. Do not substitute a default password.

load_cql_credentials() {
  CQL_USER=$(jq -er '.cassandra.dbUser.user | strings | select(length > 0)' \
    "$BACKUP_DIR/cassandra-values.json")
  CQL_PASSWORD=$(jq -er '.data["cassandra-password"] | @base64d | select(length > 0)' \
    "$BACKUP_DIR/cassandra-secret.json")
}
CQL_POD=cassandra-0
cql() {
  printf '%s\n%s\n' "$CQL_USER" "$CQL_PASSWORD" |
    kcp -n cassandra-system exec -i "$CQL_POD" -c cassandra -- \
      sh -c 'IFS= read -r u; IFS= read -r p; cqlsh -u "$u" -p "$p" "$@"' \
      -- --no-color -e "$1"
}
cql_json() {
  local output=$1 query=$2
  cql "PAGING OFF; CONSISTENCY LOCAL_QUORUM; $query" > "${output}.txt"
  python3 "$CHECKER" cql-json "${output}.txt" > "$output"
}
save_session
load_cql_credentials
cql 'DESCRIBE SCHEMA;' > "$BACKUP_DIR/schema-before.cql"
cql_json "$BACKUP_DIR/keyspaces-before.json" \
  'SELECT JSON keyspace_name, replication FROM system_schema.keyspaces;'
jq -e 'length > 0' "$BACKUP_DIR/keyspaces-before.json" >/dev/null

Keep the saved credentials private and this shell open. They remain available after the old Secret is removed. The password travels over stdin to the pod, not in the kubectl command arguments. Do not enable shell tracing.

Preserve the target credentials

In the existing target secrets/$HELMFILE_ENV-secrets.yaml, set cassandra.dbUser.user and cassandra.dbUser.password to the recorded superuser values. Set cassandra.serviceRolePassword to the source Helm value with that name. Preserve the existing openbao.migrations.env entry named DEFAULT_CASSANDRA_PASSWORD and its other array entries. This entry must equal the service-role password, which can differ from the superuser password.

cassandra:
  dbUser:
    user: "existing-superuser-name"
    password: "existing-superuser-password"
  serviceRolePassword: "existing-service-role-password"

Merge these keys into the existing secrets file; do not create a second cassandra key or replace the file. In 0.6.1, the environment file forwards persistence.subPath, but does not forward dbUser, serviceRolePassword, hooks, cluster, or config. Those chart settings belong in this secrets values file. A selected target render is checked before deleting the StatefulSet.

Record a nonempty data and migration baseline

Pause changes to the account records during this maintenance window. The example uses the normally small nvcf_api.accounts table; avoid unbounded counts on large workload tables. A missing or empty table cannot serve as a survival baseline. Record an existing function and its successful invocation through your normal client as an additional application-level baseline.

capture_data() {
  local directory=$1
  mkdir -p "$directory"
  cql_json "$directory/accounts-count.json" 'SELECT JSON count(*) FROM nvcf_api.accounts;'
  cql_json "$directory/accounts.json" 'SELECT JSON nca_id, name FROM nvcf_api.accounts;'
  jq -e --slurpfile rows "$directory/accounts.json" \
    'length == 1 and .[0].count > 0 and .[0].count == ($rows[0] | length)' \
    "$directory/accounts-count.json" >/dev/null
  cql_json "$directory/migration-tables.json" \
    "SELECT JSON table_name FROM system_schema.tables WHERE keyspace_name = 'schema_migrations';"
  jq -e 'length > 0 and all(.[]; .table_name | test("^[a-z][a-z0-9_]*$"))' \
    "$directory/migration-tables.json" >/dev/null
  local tables table
  tables=$(jq -r '.[].table_name' "$directory/migration-tables.json")
  for table in $tables; do
    cql_json "$directory/migration-$table.json" \
      "SELECT JSON version, dirty FROM schema_migrations.$table;"
  done
}
save_session
capture_data "$BACKUP_DIR/data-before"
python3 "$CHECKER" ledger "$BACKUP_DIR/data-before" > "$BACKUP_DIR/ledger-before.json"

Every discovered migration table must contain exactly one clean version record. The checker rejects CQL errors, partial output, empty inventories, missing or dirty records, and later missing tables or regressed versions. Target migrations can advance versions or add tables. Do not clear dirty flags to bypass a failure.

Back up every Cassandra member

Keep application writes paused until the baseline and snapshots are complete. Snapshots are node-local; copying one node is not a cluster backup. These commands save every tagged snapshot, schema, credentials, configuration, and identity evidence outside the cluster. Confirm that the destination satisfies your site’s backup retention and access controls.

The capacity check conservatively requires one current data footprint free on each node before flushing, and space for all nodes plus 10 percent at the backup destination. Resolve insufficient capacity before taking snapshots.

for pod in cassandra-0 cassandra-1 cassandra-2; do
  kcp -n cassandra-system exec "$pod" -c cassandra -- sh -ec '
    test -d /bitnami/cassandra/data/data
    command -v tar >/dev/null
    du -sk /bitnami/cassandra/data/data
    df -Pk /bitnami/cassandra/data/data
  ' > "$BACKUP_DIR/$pod-space.txt"
  kcp -n cassandra-system exec "$pod" -c cassandra -- \
    cat /opt/bitnami/cassandra/conf/cassandra.yaml > "$BACKUP_DIR/$pod-cassandra.yaml"
done
python3 "$CHECKER" space "$BACKUP_DIR"
SNAPSHOT_TAG="pre-0.6.1-$(date -u +%Y%m%dT%H%M%SZ)"
save_session
printf '%s\n' "$SNAPSHOT_TAG" > "$BACKUP_DIR/snapshot-tag.txt"
for pod in cassandra-0 cassandra-1 cassandra-2; do
  capture_members "$BACKUP_DIR/pre-snapshot-$pod"
  python3 "$CHECKER" topology "$BACKUP_DIR/pre-snapshot-$pod" \
    --baseline "$BACKUP_DIR/topology-before.json" --unchanged >/dev/null
  assert_context
  kcp -n cassandra-system exec "$pod" -c cassandra -- nodetool flush
  kcp -n cassandra-system exec "$pod" -c cassandra -- nodetool snapshot -t "$SNAPSHOT_TAG"
  kcp -n cassandra-system exec "$pod" -c cassandra -- nodetool listsnapshots \
    > "$BACKUP_DIR/$pod-snapshots.txt"
  kcp -n cassandra-system exec "$pod" -c cassandra -- sh -ec '
    cd /bitnami/cassandra/data
    paths=$(find data -type d -path "*/snapshots/$1")
    test -n "$paths"
    printf "%s\n" "$paths" | tar -czf - -T -
  ' -- "$SNAPSHOT_TAG" > "$BACKUP_DIR/$pod-$SNAPSHOT_TAG.tar.gz"
  python3 "$CHECKER" archive "$BACKUP_DIR/$pod-$SNAPSHOT_TAG.tar.gz" "$SNAPSHOT_TAG" \
    > "$BACKUP_DIR/$pod-backup.json"
done

Each archive must be readable to its end, contain the requested tag, include application, authentication, and migration-state snapshots, and contain the SSTables listed in its manifests, including their data, metadata, compression or checksum information, and primary index files. A table-of-contents file is required for SSTables with SAI components. It is optional only for SSTables without SAI components, as seen in released system-table snapshots. An SAI column index can omit data components only when its completion marker identifies an empty index. The checker validates the released marker format and rejects missing primary data or partial nonempty indexes. See the Cassandra SAI index format. The checker records a SHA-256 for each archive. A checksum and readable archive establish transfer integrity, not a successful restore. Keep a separately rehearsed recovery procedure. If your backup destination is temporary, copy the whole evidence directory to durable protected storage and compare the archive checksums there before proceeding. Never remove the only backup, clear snapshots, or delete PVCs as an automatic retry.

Choose a migration method

Choose the method that fits your deployment and downtime requirements. The commands above and the in-place adoption checks cover the standard three-member source cluster. Datacenter expansion and backup/restore remain outlines that need a separate site-tested execution plan; they are not qualified by the in-place checks.

Datacenter expansion keeps the cluster online by adding the new stack as a second Cassandra datacenter and streaming data to it, then cutting clients over. It has the most moving parts and needs enough capacity to run both datacenters during the migration.

In-place volume adoption reuses the existing data volumes. It is the smallest change in footprint, but it requires deleting and recreating the Cassandra StatefulSet (its volumes are retained) and pointing the new pods at the existing Bitnami on-disk layout.

Backup and restore is the most conservative for data integrity: the existing cluster is left untouched until a fresh 0.6.1 cluster is verified. It needs a maintenance window and capacity to run both clusters during the copy.

Method: datacenter expansion

Add the 0.6.1 stack as a second Cassandra datacenter in the same cluster, replicate the data to it, move clients over, then remove the old datacenter. Because the new nodes bootstrap empty and receive their data by streaming, this method avoids the volume-layout and configuration-compatibility concerns of an in-place adoption.

1. Stand up the new datacenter joined to the existing cluster. Deploy the 0.6.1 Cassandra nodes with the same cluster name as the existing cluster, a distinct datacenter name, seeds that reach the existing datacenter, and auto_bootstrap: false so the new nodes join without streaming during bootstrap. Do not run the schema-initialization or migration jobs yet. The schema already exists in the cluster, and the new nodes receive it through gossip. Disable both jobs in the new datacenter’s values with cassandra.hooks.initializeCluster.enabled: false and cassandra.hooks.migrations.enabled: false. Keep them disabled until the new datacenter is included in the application and migration-state keyspaces in step 3.

2. Confirm both datacenters are visible and healthy from a node in either datacenter:

   kcp -n cassandra-system exec cassandra-0 -c cassandra -- nodetool status

   Every node should report UN. The new datacenter nodes appear with no or minimal load until the rebuild in step 4.

3. Extend replication for every application keyspace, schema_migrations, system_auth, system_distributed, and system_traces to include the new datacenter. Replicating schema_migrations preserves the version and dirty state that the 0.6.1 migration job uses to apply only pending migrations. For each keyspace:

   ALTER KEYSPACE <keyspace> WITH replication = {
     'class': 'NetworkTopologyStrategy',
     '<old-datacenter>': <rf>,
     '<new-datacenter>': <rf>
   };

4. On each node in the new datacenter, stream the existing data across:

   kcp -n cassandra-system exec "<new-dc-pod>" -c cassandra -- \
     nodetool rebuild -- "<old-datacenter>"

5. Run the pending 0.6.1 schema migrations before moving clients. Keep cluster initialization disabled and enable only the migration hook:

   cassandra:
     hooks:
       initializeCluster:
         enabled: false
       migrations:
         enabled: true

   Apply the Cassandra release and wait for its migration job:

   set -euo pipefail
   assert_context
   make install HELMFILE_ENV="$HELMFILE_ENV" \
     KUBECONFIG_FILE="$CONTROL_KUBECONFIG" HELMFILE_SELECTOR=name=cassandra
   kcp -n cassandra-system wait \
     --for=condition=complete job/cassandra-migrations --timeout=10m
   kcp -n cassandra-system logs job/cassandra-migrations

   The logs must end with All Cassandra migrations completed successfully. Do not continue if the job failed or timed out.

6. Inventory application keyspace replication after the migration job:

   SELECT keyspace_name, replication
   FROM system_schema.keyspaces;

   For each application keyspace that does not list both datacenters, update its replication:

   ALTER KEYSPACE <keyspace> WITH replication = {
     'class': 'NetworkTopologyStrategy',
     '<old-datacenter>': <rf>,
     '<new-datacenter>': <rf>
   };

   Run the system_schema.keyspaces query again. Confirm that every application keyspace lists both datacenters before continuing.

   If a newly created keyspace already contains data, rebuild that keyspace on each new-datacenter node before continuing:

   kcp -n cassandra-system exec "<new-dc-pod>" -c cassandra -- \
     nodetool rebuild -ks "<keyspace>" -- "<old-datacenter>"

7. Confirm the migration state is clean:

   SELECT table_name
   FROM system_schema.tables
   WHERE keyspace_name = 'schema_migrations';
   SELECT version, dirty FROM schema_migrations.<table-name>;

   Run the second query for each table returned by the first query. Each query must return one record with dirty set to false. This confirms that the cluster can receive later schema migrations. Do not move clients or remove the old datacenter while a migration is missing or dirty.

8. Move clients to the new datacenter. Point application traffic at the new datacenter using LOCAL_QUORUM against the new datacenter name, and confirm reads and writes succeed there.

9. Remove the old datacenter from replication, then decommission it. Alter every keyspace again to drop the old datacenter from the replication map, then decommission the old-datacenter nodes and remove the old stack.

Method: in-place volume adoption

Reuse the existing data volumes with the new stack. The StatefulSet must be recreated because its shape is immutable, and the new pods must be pointed at the Bitnami on-disk layout.

Use persistence.subPath: data to expose the existing Bitnami layout at the paths the official image expects. This procedure does not relocate data files. Review the saved Cassandra configuration before continuing. Custom cluster names, datacenters, TLS, storage paths, or Cassandra settings need matching chart values and a site-tested procedure; do not assume defaults are compatible.

1. In the target environments/$HELMFILE_ENV.yaml, merge this value into the existing cassandra section:

   cassandra:
     persistence:
       subPath: "data"

2. Render the site’s target Cassandra and OpenBao charts before stopping any old pod. Keep this output private because it contains credentials. The checker verifies the replica count, subPath, cluster/datacenter names, released runtime overrides, and preserved superuser/service credentials. The OpenBao render verifies the shared service credential; it does not apply or restart OpenBao.

   cd "$TARGET_CONTROL"
   assert_context
   python3 "$CHECKER" bundle "$TARGET_CONTROL" 0.6.1
   HELMFILE_ENV="$HELMFILE_ENV" helmfile --environment default \
     --kubeconfig "$CONTROL_KUBECONFIG" \
     --selector name=cassandra --selector name=openbao-server template \
     > "$BACKUP_DIR/target-selected.yaml"
   kcp create --dry-run=client --validate=false -f "$BACKUP_DIR/target-selected.yaml" \
     -o json > "$BACKUP_DIR/target-selected.json"
   python3 "$CHECKER" target "$BACKUP_DIR/target-selected.json" \
     "$BACKUP_DIR/cassandra-values.json" "$BACKUP_DIR/cassandra-secret.json" \
     "$BACKUP_DIR/before/statefulset.json" "$BACKUP_DIR" > "$BACKUP_DIR/target-checked.json"
   TARGET_CASSANDRA_IMAGE=$(jq -er '.image' "$BACKUP_DIR/target-checked.json")
   save_session

3. Recheck the original members and every off-node archive immediately before adoption. Require a fresh, clean ledger and the same account baseline.

   capture_members "$BACKUP_DIR/pre-adoption"
   python3 "$CHECKER" topology "$BACKUP_DIR/pre-adoption" \
     --baseline "$BACKUP_DIR/topology-before.json" --unchanged >/dev/null
   capture_data "$BACKUP_DIR/data-pre-adoption"
   python3 "$CHECKER" ledger "$BACKUP_DIR/data-pre-adoption" \
     --baseline "$BACKUP_DIR/ledger-before.json" >/dev/null
   diff -u "$BACKUP_DIR/data-before/accounts.json" "$BACKUP_DIR/data-pre-adoption/accounts.json"
   diff -u "$BACKUP_DIR/data-before/accounts-count.json" "$BACKUP_DIR/data-pre-adoption/accounts-count.json"
   for pod in cassandra-0 cassandra-1 cassandra-2; do
     python3 "$CHECKER" archive "$BACKUP_DIR/$pod-$SNAPSHOT_TAG.tar.gz" "$SNAPSHOT_TAG" \
       > "$BACKUP_DIR/$pod-backup-recheck.json"
     cmp "$BACKUP_DIR/$pod-backup.json" "$BACKUP_DIR/$pod-backup-recheck.json"
   done
   assert_context
   test "$(kcp -n cassandra-system get statefulset cassandra -o jsonpath='{.metadata.uid}')" \
     = "$(jq -r '.statefulset_uid' "$BACKUP_DIR/topology-before.json")"
   kcp -n cassandra-system delete statefulset cassandra --cascade=orphan
   assert_context
   test "$(kcp -n cassandra-system get pod cassandra-0 -o jsonpath='{.metadata.uid}')" \
     = "$(jq -r '.members["cassandra-0"].uid' "$BACKUP_DIR/topology-before.json")"
   kcp -n cassandra-system delete pod cassandra-0
   assert_context
   make install HELMFILE_ENV="$HELMFILE_ENV" \
     KUBECONFIG_FILE="$CONTROL_KUBECONFIG" HELMFILE_SELECTOR=name=cassandra
   kcp -n cassandra-system rollout status statefulset/cassandra --timeout=30m
   kcp -n cassandra-system wait --for=condition=complete \
     job/cassandra-initialize-cluster job/cassandra-migrations --timeout=10m
   kcp -n cassandra-system get jobs -o json > "$BACKUP_DIR/jobs-after-cassandra.json"
   python3 "$CHECKER" jobs "$BACKUP_DIR/jobs-after-cassandra.json" \
     "$BACKUP_DIR/jobs-before.json"
   kcp -n cassandra-system logs job/cassandra-migrations \
     > "$BACKUP_DIR/cassandra-migrations.log"

   The StatefulSet controller recreates cassandra-0 and replaces the remaining pods in reverse ordinal order. Do not delete those pods manually. A failed install or timeout stops the procedure even if a pod later becomes Ready. Keep the PVCs and backups and inspect the ring, events, and hook logs.

4. Continue to Verify Cassandra. A new Ready pod or a successful Helm command alone does not establish adoption.

Method: backup and restore

Leave the existing cluster in place and load its data into a fresh 0.6.1 cluster.

1. Snapshot the existing cluster: on every node, run nodetool flush then nodetool snapshot, and copy the snapshots off-node.

2. Deploy a fresh 0.6.1 Cassandra cluster on new volumes and let the init and migration jobs create the schema.

3. Restore the data into the new cluster with sstableloader, or place the snapshot sstables and run nodetool refresh per table.

4. Verify row counts and application connectivity against the new cluster, then decommission the old cluster.

Verify Cassandra before continuing

For in-place adoption, require all three new pods on the target image and StatefulSet revision, the same original PVC UIDs/PVs and Cassandra host IDs, and the same healthy ring. All original pods must have been replaced. For a separately tested datacenter or restore migration, use that procedure’s identity mapping instead; the in-place identity comparison does not apply to new volumes.

capture_members "$BACKUP_DIR/after-cassandra"
python3 "$CHECKER" topology "$BACKUP_DIR/after-cassandra" \
  --baseline "$BACKUP_DIR/topology-before.json" --image "$TARGET_CASSANDRA_IMAGE" \
  > "$BACKUP_DIR/topology-after-cassandra.json"
capture_data "$BACKUP_DIR/data-after-cassandra"
python3 "$CHECKER" ledger "$BACKUP_DIR/data-after-cassandra" \
  --baseline "$BACKUP_DIR/ledger-before.json" > "$BACKUP_DIR/ledger-after-cassandra.json"
diff -u "$BACKUP_DIR/data-before/accounts.json" "$BACKUP_DIR/data-after-cassandra/accounts.json"
diff -u "$BACKUP_DIR/data-before/accounts-count.json" "$BACKUP_DIR/data-after-cassandra/accounts-count.json"

Do not proceed if these commands fail. Count equality is a bounded baseline check, not proof that every table survived. Confirm that the original function and its deployment are still retrievable through your existing client.

Apply the remaining 0.6.1 stack

The Cassandra selector leaves other releases at their old versions. After Cassandra verification, render the full target stack with the same site files and review the changes. The template contains secrets; keep it in the private backup directory. Clear an inherited selector explicitly on the full install.

cd "$TARGET_CONTROL"
assert_context
python3 "$CHECKER" bundle "$TARGET_CONTROL" 0.6.1
HELMFILE_ENV="$HELMFILE_ENV" helmfile --environment default \
  --kubeconfig "$CONTROL_KUBECONFIG" template > "$BACKUP_DIR/target-full.yaml"
kcp -n cassandra-system get jobs -o json > "$BACKUP_DIR/jobs-before-full.json"

Check the render against your source configuration before running the next block. Preserve the Cassandra subPath and credentials and OpenBao OnDelete strategy. Do not restart OpenBao at this stage. Save the existing OpenBao pod UIDs and live images for the required continuation below.

kcp -n vault-system get pods -o json > "$BACKUP_DIR/openbao-before-full.json"
assert_context
make install HELMFILE_ENV="$HELMFILE_ENV" \
  KUBECONFIG_FILE="$CONTROL_KUBECONFIG" HELMFILE_SELECTOR=
HELMFILE_ENV="$HELMFILE_ENV" helmfile --environment default \
  --kubeconfig "$CONTROL_KUBECONFIG" list --skip-charts --output json \
  > "$BACKUP_DIR/expected-releases.json"
helm --kubeconfig "$CONTROL_KUBECONFIG" --kube-context "$CONTROL_CONTEXT" \
  list --all-namespaces --all --max 10000 --output json > "$BACKUP_DIR/installed-releases.json"
python3 "$CHECKER" releases "$BACKUP_DIR/expected-releases.json" \
  "$BACKUP_DIR/installed-releases.json"
kcp -n cassandra-system get jobs -o json > "$BACKUP_DIR/jobs-after-full.json"
python3 "$CHECKER" jobs "$BACKUP_DIR/jobs-after-full.json" "$BACKUP_DIR/jobs-before-full.json"
capture_members "$BACKUP_DIR/after-full"
python3 "$CHECKER" topology "$BACKUP_DIR/after-full" \
  --baseline "$BACKUP_DIR/topology-before.json" --image "$TARGET_CASSANDRA_IMAGE" >/dev/null
capture_data "$BACKUP_DIR/data-after-full"
python3 "$CHECKER" ledger "$BACKUP_DIR/data-after-full" \
  --baseline "$BACKUP_DIR/ledger-after-cassandra.json" >/dev/null
diff -u "$BACKUP_DIR/data-before/accounts.json" "$BACKUP_DIR/data-after-full/accounts.json"
diff -u "$BACKUP_DIR/data-before/accounts-count.json" "$BACKUP_DIR/data-after-full/accounts-count.json"

The release comparison requires every enabled target release to be deployed with its expected chart version. It includes optional releases when enabled; it does not compare unrelated Helm releases. Recheck application readiness, the original function, and fresh authentication. This command updates the control-plane bundle, not the separate compute-plane or observability bundles.

Continue through the OpenBao 0.6.2 patch

The full 0.6.1 apply updates the OpenBao StatefulSet template to 2.5.5-nv-1.3.1. Its OnDelete strategy leaves existing server pods running 2.5.4-nv-1.3.0 until they are replaced. A deployed Helm release does not mean those running servers have upgraded.

Compare the template and the actual server and auto-unseal container images:

kcp -n vault-system get statefulset openbao-server \
  -o jsonpath='{.spec.updateStrategy.type}{"\n"}{range .spec.template.spec.containers[*]}{.name}{" "}{.image}{"\n"}{end}'
kcp -n vault-system get pods -l app.kubernetes.io/name=openbao,component=server \
  -o json | jq -r '.items[] | .metadata.name as $pod |
    .metadata.labels["controller-revision-hash"] as $revision |
    .spec.containers[] | [$pod,.name,.image,$revision] | @tsv'

Do not restart or delete OpenBao server pods, reload the JWT plugin, or change the update strategy at 0.6.1. The old plugin catalog must be refreshed before new binaries can start safely. Keep the healthy existing members running and continue directly to the 0.6.1 to 0.6.2 procedure, which detects inherited old pods even when the target template is unchanged, refreshes the catalog, and replaces standbys before gracefully replacing the active member.

If a member has already restarted, is sealed, or cannot issue fresh JWTs, stop and assess that state before continuing. Do not delete another member to recover. The overall upgrade is complete only after the 0.6.2 runtime, per-pod revision, Raft, authentication and invocation checks pass.