Merge pull request #2166 from AlexsLemonade/dev

[DEPLOY] Deploy final fixes to pg_bouncer config and fix intermittent test failures.

common/data_refinery_common/models/original_file.py

@@ -230,6 +230,14 @@ def needs_downloading(self, own_processor_id=None) -> bool:
         if unstarted_downloader_jobs.count() > 0:
             return False
 
+        # Do an extra check for blocking jobs for trancsriptome indices.
+        # This is necessary because needs_processing() won't check
+        # the blocking jobs for them because they're supposed to
+        # have multiple processor jobs. However if the file does need to
+        # be redownloaded, we only want one downloader job to be recreated.
+        if self.has_blocking_jobs(own_processor_id):
+            return False
+
         # If this file has been processed, then it doesn't need to be downloaded again.
         return self.needs_processing(own_processor_id)
 

infrastructure/deploy.sh

@@ -257,14 +257,14 @@ docker pull $DOCKERHUB_REPO/$FOREMAN_DOCKER_IMAGE
 # Test that the pg_bouncer instance is up. 15 minutes should be more than enough.
 start_time=$(date +%s)
 diff=0
-until pg_isready -d $DATABASE_NAME -h $DATABASE_HOST -p $DATABASE_PORT -U $DATABASE_USER &> /dev/null || [ "$diff" -gt "900" ]
+until pg_isready -d $DATABASE_NAME -h $DATABASE_PUBLIC_HOST -p $DATABASE_PORT -U $DATABASE_USER &> /dev/null || [ "$diff" -gt "900" ]
 do
     echo "Waiting for the pg_bouncer instance to come online ..."
     sleep 10
     let "diff = $(date +%s) - $start_time"
 done
 
-if ! pg_isready -d $DATABASE_NAME -h $DATABASE_HOST -p $DATABASE_PORT -U $DATABASE_USER &> /dev/null; then
+if ! pg_isready -d $DATABASE_NAME -h $DATABASE_PUBLIC_HOST -p $DATABASE_PORT -U $DATABASE_USER &> /dev/null; then
     echo "pg_bouncer instance failed to come up after 15 minutes."
     exit 1
 fi
@@ -273,18 +273,21 @@ fi
 docker run \
        --env-file prod_env \
        --env RUNNING_IN_CLOUD=False \
+       --env DATABASE_HOST=$DATABASE_PUBLIC_HOST \
        $DOCKERHUB_REPO/$FOREMAN_DOCKER_IMAGE python3 manage.py migrate auth
 
 # Apply general migrations.
 docker run \
        --env-file prod_env \
        --env RUNNING_IN_CLOUD=False \
+       --env DATABASE_HOST=$DATABASE_PUBLIC_HOST \
        $DOCKERHUB_REPO/$FOREMAN_DOCKER_IMAGE python3 manage.py migrate
 
 # Create the cache table if it does not already exist.
 docker run \
        --env-file prod_env \
        --env RUNNING_IN_CLOUD=False \
+       --env DATABASE_HOST=$DATABASE_PUBLIC_HOST \
        $DOCKERHUB_REPO/$FOREMAN_DOCKER_IMAGE python3 manage.py createcachetable
 
 # Make sure to clear out any old nomad job specifications since we

infrastructure/variables.tf

@@ -194,6 +194,10 @@ output "environment_variables" {
     {name = "DATABASE_NAME"
       value = "${aws_db_instance.postgres_db.name}"},
     {name = "DATABASE_HOST"
+      value = "${aws_instance.pg_bouncer.private_ip}"},
+    # We want to use the private IP from everywhere except wherever
+    # deployment is happening, so we also need to expose this IP.
+    {name = "DATABASE_PUBLIC_HOST"
       value = "${aws_instance.pg_bouncer.public_ip}"},
     {name = "RDS_HOST"
       value = "${aws_db_instance.postgres_db.address}"},

workers/data_refinery_workers/processors/utils.py

@@ -31,11 +31,7 @@
     ProcessorJobOriginalFileAssociation,
     Sample,
 )
-from data_refinery_common.utils import (
-    get_env_variable,
-    get_env_variable_gracefully,
-    get_instance_id,
-)
+from data_refinery_common.utils import get_env_variable, get_instance_id
 
 logger = get_and_configure_logger(__name__)
 # Let this fail if SYSTEM_VERSION is unset.