diff --git a/bin/upgrade-scripts/to-2.0/upgrade.sh b/bin/upgrade-scripts/to-2.0/upgrade.sh
index e4fc65808ef826b9317a8390d72229eb2a609302..9399406e0a6b731d0924cc566e8c478c459dbf2f 100755
--- a/bin/upgrade-scripts/to-2.0/upgrade.sh
+++ b/bin/upgrade-scripts/to-2.0/upgrade.sh
@@ -2,11 +2,28 @@
 
 set -euo pipefail
 
-# Suspend the kube-prometheus-stack helmRelease so it cannot recreate a
-# failing v0.39 daemonset
-flux suspend helmrelease -n stackspin kube-prometheus-stack
-# Delete kube-prometheus-stack-prometheus-node-exporter daemonset, needed
-# for upgrade to kube-prometheus-stack v0.40.0
-# see https://artifacthub.io/packages/helm/prometheus-community/kube-prometheus-stack#from-39-x-to-40-x
-kubectl -n stackspin delete daemonset kube-prometheus-stack-prometheus-node-exporter
-flux resume helmrelease -n stackspin kube-prometheus-stack
+# Check if zulip is installed at all.
+if helm ls -n stackspin-apps | grep -q zulip
+then
+  zulip="true"
+else
+  zulip="false"
+fi
+
+# Prepare before upgrade.
+# For background, see
+# https://open.greenhost.net/stackspin/stackspin/-/issues/1250
+if [ $zulip == "true" ]
+then
+  flux suspend helmrelease -n stackspin-apps zulip
+  # Temporarily uninstall zulip.
+  helm uninstall -n stackspin-apps zulip
+fi
+
+# Now do the actual upgrade, by instructing flux to switch to the v2 branch.
+kubectl patch -n flux-system gitrepository stackspin -p '{"spec":{"ref":{"branch":"v2"}}}' --type="merge"
+
+if [ $zulip == "true" ]
+then
+  flux resume helmrelease -n stackspin-apps zulip
+fi
diff --git a/flux2/apps/zulip/release.yaml b/flux2/apps/zulip/release.yaml
index 06ac1623e7bbd33f9c4d884a4e916fc195cce797..ba07460abdfe3d4ef9e7dde24276821cdcfdece9 100644
--- a/flux2/apps/zulip/release.yaml
+++ b/flux2/apps/zulip/release.yaml
@@ -9,7 +9,7 @@ spec:
   chart:
     spec:
       chart: zulip
-      version: 0.2.1
+      version: 0.3.0
       sourceRef:
         kind: HelmRepository
         name: zulip-helm-chart
@@ -17,9 +17,11 @@ spec:
   interval: 1h
   install:
     timeout: 30m
+    remediation:
+      retries: 1
   upgrade:
     remediation:
-      retries: 2
+      retries: 1
   valuesFrom:
     - kind: ConfigMap
       name: stackspin-zulip-values
diff --git a/flux2/apps/zulip/zulip-values-configmap.yaml b/flux2/apps/zulip/zulip-values-configmap.yaml
index 97f894aa3b1fccbee8d4bd19d25ab56d5484afb0..2efd499ffccd64227d0d34fbdbf0f61138cad1ed 100644
--- a/flux2/apps/zulip/zulip-values-configmap.yaml
+++ b/flux2/apps/zulip/zulip-values-configmap.yaml
@@ -119,9 +119,14 @@ data:
     postgresql:
       image:
         tag: 14
-      persistence:
-        existingClaim: zulip-postgres
-      postgresqlPassword: "${postgresql_password}"
+      auth:
+        # Note: the old version (< 0.3.0) of this chart always used the
+        # postgresql admin password to log in. since 0.3.0 it uses a user
+        # password, but because we can't update the passwords in currently
+        # running postgresql instances through helm, we use the same password
+        # for both.
+        postgresqlPassword: "${postgresql_password}"
+        password: "${postgresql_password}"
       resources:
         limits:
           memory: 512Mi
@@ -129,6 +134,8 @@ data:
           cpu: 10m
           memory: 256Mi
       primary:
+        persistence:
+          existingClaim: zulip-postgres
         podAnnotations:
           backup.velero.io/backup-volumes: "data"
       commonLabels: