Commit 3dca7c0

apiVersion: apps.cozystack.io/v1alpha1

kind: FoundationDB

metadata:

name: backup-contract

spec:

cluster:

version: "7.3.63"

processCounts:

storage: 1

stateless: -1

cluster_controller: 1

redundancyMode: "single"

storageEngine: "ssd-2"

faultDomain:

key: "foundationdb.org/none"

valueFrom: "$FDB_ZONE_ID"

storage:

size: "1Gi"

storageClass: ""

resourcesPreset: "small"

monitoring:

enabled: false

customParameters:

- "knob_disable_posix_kernel_aio=1"

imageType: "unified"

automaticReplacements: true

set -euo pipefail

: "${NAMESPACE:?NAMESPACE must be set}"

cd "$(dirname "${BASH_SOURCE[0]}")/../../.."

log_substep() { echo " -> $*"; }

log_success() { echo " OK $*"; }

log_error() { echo " ERROR $*" >&2; }

. examples/backups/_lib/wait-helpers.sh

SRC=fdb-rt-src

DST=fdb-rt-copy

BJ1=fdb-rt-bj1

BJ2=fdb-rt-bj2

RJ=fdb-rt-restore

SENTINEL_KEY=/e2e/fdb-roundtrip

OWNED_BY=backups.cozystack.io/owned-by.BackupJobName

apply_app() {

kubectl -n "$NAMESPACE" apply -f - <<EOF

}

apply_backupjob() {

kubectl -n "$NAMESPACE" apply -f - <<EOF

}

fdbcli_exec() {

local pod

pod=$(kubectl -n "$NAMESPACE" get pods \

[[ -n "$pod" ]] || { log_error "no running cluster_controller pod for foundationdb-$1"; return 1; }

timeout -k 5 60 kubectl -n "$NAMESPACE" exec "$pod" -c foundationdb -- \

fdbcli --cluster-file=/var/dynamic-conf/fdb.cluster --exec "$2"

}

backup_uri() {

kubectl -n "$NAMESPACE" get backup "$1" -o json | jq -r '.spec.driverMetadata["apps.foundationdb.org/storage-uri"] // ""'

}

describe_dir() {

local deploy out rc=0

deploy=$(kubectl -n "$NAMESPACE" get deploy -o name | grep -- "^deployment.apps/$BJ2-.*-backup-agents$" | head -n1)

[[ -n "$deploy" ]] || { log_error "no backup agents Deployment for $BJ2"; return 2; }

out=$(timeout -k 5 120 kubectl -n "$NAMESPACE" exec "$deploy" -c foundationdb -- fdbbackup describe -d "$1" 2>&1) || rc=$?

if [[ $rc -eq 0 ]]; then

echo present

elif grep -q 'Backup does not exist' <<<"$out"; then

echo absent

else

log_error "fdbbackup describe failed for $1: $out"

return 2

fi

}

run() {

local value backup1 backup2 uri1 uri2 state1 state2 got

value="roundtrip-$(date -u +%Y%m%dT%H%M%SZ)"

apply_app "$SRC"

apply_app "$DST"

# 660s: above the 600s install timeout cozystack-api gives these releases

# (no helm-install-timeout override on foundationdb-rd), per the HR-Ready

# sizing in docs/agents/e2e-testing.md.

wait_hr_ready "foundationdb-$SRC" 660

wait_hr_ready "foundationdb-$DST" 660

wait_for_field foundationdbclusters.apps.foundationdb.org "foundationdb-$SRC" '{.status.health.available}' true "$NAMESPACE" 600

wait_for_field foundationdbclusters.apps.foundationdb.org "foundationdb-$DST" '{.status.health.available}' true "$NAMESPACE" 600

fdbcli_exec "$SRC" "writemode on; set $SENTINEL_KEY $value"

# A BackupJob succeeds once its own directory is restorable; 900s covers

# agent start, the first snapshot and the driver's 60s status refresh.

apply_backupjob "$BJ1"

wait_for_field backupjob "$BJ1" '{.status.phase}' Succeeded "$NAMESPACE" 900 Failed

apply_backupjob "$BJ2"

wait_for_field backupjob "$BJ2" '{.status.phase}' Succeeded "$NAMESPACE" 900 Failed

backup1=$(kubectl -n "$NAMESPACE" get backupjob "$BJ1" -o jsonpath='{.status.backupRef.name}')

backup2=$(kubectl -n "$NAMESPACE" get backupjob "$BJ2" -o jsonpath='{.status.backupRef.name}')

uri1=$(backup_uri "$backup1")

uri2=$(backup_uri "$backup2")

[[ "$uri1" == *"/$NAMESPACE/$BJ1?"* ]] || { log_error "first backup does not name its own directory: $uri1"; return 1; }

[[ "$uri2" == *"/$NAMESPACE/$BJ2?"* ]] || { log_error "second backup does not name its own directory: $uri2"; return 1; }

kubectl -n "$NAMESPACE" apply -f - <<EOF

wait_for_field restorejob "$RJ" '{.status.phase}' Succeeded "$NAMESPACE" 900 Failed

got=$(fdbcli_exec "$DST" "get $SENTINEL_KEY")

if ! grep -qF "is \`$value'" <<<"$got"; then

log_error "sentinel did not survive the restore: $got"

return 1

fi

log_success "sentinel restored into $DST"

kubectl -n "$NAMESPACE" delete backup "$backup1" --wait=false

wait_deleted backup "$backup1" 600

state1=$(describe_dir "$uri1")

state2=$(describe_dir "$uri2")

[[ "$state1" == absent ]] || { log_error "deleting the first Backup left its directory: $uri1"; return 1; }

[[ "$state2" == present ]] || { log_error "deleting the first Backup removed the second's directory: $uri2"; return 1; }

log_success "FoundationDB round-trip verified"

}

cleanup() {

kubectl -n "$NAMESPACE" delete restorejob "$RJ" --ignore-not-found --wait=false

# Each Backup's finalizer stops its operator CR and deletes its directory.

local bj b

for bj in "$BJ1" "$BJ2"; do

b=$(kubectl -n "$NAMESPACE" get backupjob "$bj" --ignore-not-found -o jsonpath='{.status.backupRef.name}')

if [[ -n "$b" ]]; then

kubectl -n "$NAMESPACE" delete backup "$b" --ignore-not-found --wait=false

wait_deleted backup "$b" 600

fi

done

kubectl -n "$NAMESPACE" delete backupjob "$BJ1" "$BJ2" --ignore-not-found --wait=false

kubectl -n "$NAMESPACE" delete foundationdbbackups.apps.foundationdb.org -l "$OWNED_BY in ($BJ1,$BJ2)" --ignore-not-found

kubectl -n "$NAMESPACE" delete foundationdbrestores.apps.foundationdb.org -l "$OWNED_BY=$RJ" --ignore-not-found

kubectl -n "$NAMESPACE" delete foundationdbs.apps.cozystack.io "$SRC" "$DST" --ignore-not-found --wait=false

wait_deleted hr "foundationdb-$SRC" 600

wait_deleted hr "foundationdb-$DST" 600

}

case "${1:-}" in

run) run ;;

cleanup) cleanup ;;

*) echo "usage: $0 run|cleanup" >&2; exit 2 ;;