diff --git a/docs/guides/benchmarking.md b/docs/guides/benchmarking.md
index 829ef4d9fc..42bc6455e4 100644
--- a/docs/guides/benchmarking.md
+++ b/docs/guides/benchmarking.md
@@ -780,8 +780,10 @@ what to set before the run:
   timeout. Three misses of the liveness probe (`pg_isready`) restart
   postgres mid-run; three misses of the readiness probe (`nc`) drop the pod
   from the Service, and ateapi loses its connections without a restart. Give
-  both `timeoutSeconds: 5` and `failureThreshold: 6`, and give the postgres
-  container CPU and memory requests sized for the run: on a saturated node
+  the readiness probe `timeoutSeconds: 5` and the liveness probe
+  `timeoutSeconds: 10`, so a slow database leaves the Service before it is
+  killed, both with `failureThreshold: 6`, and give the postgres container
+  CPU and memory requests sized for the run: on a saturated node
   the request is its share of the CPU, and a starved postgres is what misses
   the probes.
 * **Parking budget.** `--parked-request-budget` (default `5s`) has to cover a
diff --git a/manifests/ate-install/postgres/postgres.yaml b/manifests/ate-install/postgres/postgres.yaml
index 7e41652b6b..9364f20a90 100644
--- a/manifests/ate-install/postgres/postgres.yaml
+++ b/manifests/ate-install/postgres/postgres.yaml
@@ -183,6 +183,8 @@ spec:
             - nc -z 127.0.0.1 5432
           initialDelaySeconds: 2
           periodSeconds: 2
+          timeoutSeconds: 5
+          failureThreshold: 6
         livenessProbe:
           exec:
             command:
@@ -191,6 +193,8 @@ spec:
             - pg_isready -U "${POSTGRES_USER}" -d postgres
           initialDelaySeconds: 10
           periodSeconds: 10
+          timeoutSeconds: 10
+          failureThreshold: 6
         args:
         - -c
         - config_file=/etc/postgresql/postgresql.conf