Test upgrade of 2023.06

LAVA from 2023.01 to 2023.05 miss a migration script which is generated by
lava-docker but with a "automatic" name.
LAVA 2023.06 has the correct migration script but you cannot have both
at the same time.
Add a detection of this situation to rollback automatic migration.

In the same time replace all migration action via the postinst.py which
do the correct action.

Signed-off-by: Corentin LABBE <[email protected]>

Author: montjoie

.github/workflows/basic.yml

@@ -80,3 +80,21 @@ jobs:
         run: cd output/local && docker-compose up -d
       - name: Wait for LAVA to be started
         run: sh .github/workflows/wait-for-docker.sh
+
+      - name: Run backup of 2023.01
+        run: ./backup.sh
+      - name: stop docker
+        run: cd output/local && docker-compose down
+
+      - name: Clean old install
+        run: rm -r output
+      - name: Copy backup
+        run: cp -v backup-latest/* lava-master/backup/
+      - name: Run lavalab-gen
+        run: ./lavalab-gen.py boards-ci-2023.06.yaml
+      - name: Build lava-docker 2023.06
+        run: cd output/local && docker-compose build
+      - name: Launch lava-docker 2023.06
+        run: cd output/local && docker-compose up -d
+      - name: Wait for LAVA 2023.06 to be started
+        run: sh .github/workflows/wait-for-docker.sh

.github/workflows/wait-for-docker.sh

@@ -22,7 +22,17 @@ do
 			fi
 		fi
 	fi
-	docker-compose logs --tail=50
+	docker-compose logs --tail=60
+	docker ps > /tmp/alldocker
+	grep -q master /tmp/alldocker
+	if [ $? -ne 0 ];then
+		echo "=========================================="
+		echo "=========================================="
+		echo "=========================================="
+		echo "ERROR: master died"
+		docker-compose logs
+		exit 1
+	fi
 	sleep 10
 	TIMEOUT=$((TIMEOUT+10))
 done

boards-ci-2023.06.yaml

@@ -0,0 +1,29 @@
+---
+masters:
+  - name: masterci1
+    version: "2023.06"
+    host: local
+    healthcheck_url: http://healthcheck
+    users:
+      - name: admin
+        token: tokenforci
+        password: passwordforci
+        superuser: true
+        staff: true
+    tokens:
+      - username: admin
+        token: dfjdfkfkdjfkdsjfslforci
+        description: no description
+slaves:
+  - name: lab-ci-0
+    version: "2023.06"
+    host: local
+    remote_master: masterci1
+    remote_user: admin
+    use_overlay_server: false
+    use_tftp: false
+    host_healthcheck: true
+
+boards:
+  - name: qemu-01
+    type: qemu

lava-master/entrypoint.d/01_setup.sh

@@ -17,7 +17,7 @@ else
 fi
 
 cd /root/
-lava-server manage makemigrations || exit $?
+#lava-server manage makemigrations || exit $?
 
 # verify that the backup was not already applied in case of persistent_db
 if [ ! -e "/var/lib/postgresql/lava-docker.backup_done" ];then
@@ -28,7 +28,7 @@ if [ ! -e "/var/lib/postgresql/lava-docker.backup_done" ];then
 	if [ -e /root/backup/db_lavaserver ];then
 		echo "Restore database from backup"
 		sudo -u postgres psql < /root/backup/db_lavaserver || exit $?
-		yes yes | lava-server manage migrate || exit $?
+		#yes yes | lava-server manage migrate || exit $?
 		echo "Restore jobs output from backup"
 		rm -r /var/lib/lava-server/default/media/job-output/*
 
@@ -51,8 +51,58 @@ else
 	echo "DEBUG: backup already applied"
 fi
 
-lava-server manage makemigrations
-yes yes | lava-server manage migrate || exit $?
+# check current LAVA version
+# not very good way, but no real choice
+echo "DEBUG: check LAVA version from DB"
+su - postgres -c 'psql --tuples-only lavaserver -c "SELECT DISTINCT version from lava_scheduler_app_worker" | sort -V' > /tmp/workerversions
+sed -i 's,^[[:space:]]*,,' /tmp/workerversions
+echo "======"
+cat /tmp/workerversions
+echo "======"
+
+echo "DEBUG: check LAVA version from file"
+cat /usr/lib/python3/dist-packages/lava_common/VERSION
+# hack
+grep -q '2023.01' /usr/lib/python3/dist-packages/lava_common/VERSION
+if [ $? -eq 0 ];then
+	echo "DEBUG: 2023.01 need to do an old migration"
+	lava-server manage makemigrations
+	yes yes | lava-server manage migrate || exit $?
+fi
+# if we came from 2023.01 to 2023.05, we need to handle the migration bug
+grep -q '2023.0[1-5]' /tmp/workerversions
+if [ $? -eq 0 ];then
+	grep -q 2023.06 /usr/lib/python3/dist-packages/lava_common/VERSION
+	if [ $? -eq 0 ];then
+		echo "============================="
+		echo "DEBUG: handle DB migration BUG"
+		sudo -u postgres psql lavaserver -c "SELECT * from django_migrations where app = 'lava_scheduler_app';"
+		echo "============================="
+		sudo -u postgres psql lavaserver -c "SELECT * from django_migrations where app = 'lava_results_app';"
+		echo "============================="
+		lava-server manage migrate --fake lava_results_app 0019_update_query_contenttype || exit $?
+		echo "============================="
+		lava-server manage migrate --fake lava_scheduler_app 0057_dt_permissions_worker_master_version || exit $?
+		echo "============================="
+		lava-server manage migrate lava_results_app 0018_drop_buglink || exit $?
+		echo "============================="
+		lava-server manage migrate lava_scheduler_app 0056_testjob_queue_timeout || exit $?
+		echo "============================="
+		sudo -u postgres psql lavaserver -c "SELECT * from django_migrations where app = 'lava_scheduler_app';"
+		echo "============================="
+		sudo -u postgres psql lavaserver -c "SELECT * from django_migrations where app = 'lava_results_app';"
+		echo "============================="
+	else
+		echo "DEBUG: no spetial handling for DB"
+	fi
+else
+	echo "DEBUG: no spetial handling for DB"
+fi
+
+echo "DEBUG: call postinst for managing migrations"
+/usr/share/lava-server/postinst.py || exit $?
+#lava-server manage makemigrations
+#yes yes | lava-server manage migrate || exit $?
 
 # default site is set as example.com
 if [ -e /root/lava_http_fqdn ];then