[ 
https://issues.apache.org/jira/browse/AIRFLOW-2726?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Jon Davies updated AIRFLOW-2726:
--------------------------------
    Description: 
If one runs:

{code}
~/git/incubator-airflow/scripts/ci/kubernetes$ kubectl apply 
--namespace=testing-airflow -f kube/
{code}

The container fails with:

{code}
sqlalchemy.exc.OperationalError: (psycopg2.OperationalError) could not 
translate host name "postgres-airflow" to address: Name or service not known
{code}

...and then the pod tries to go into a Running state before 
CrashLoopBackOff'ing about:

{code}
sqlalchemy.exc.InvalidRequestError: This Session's transaction has been rolled 
back due to a previous exception during flush. To begin a new transaction with 
this Session, first issue Session.rollback(). Original exception was: 
(psycopg2.ProgrammingError) relation "log" does not exist
LINE 1: INSERT INTO log (dttm, dag_id, task_id, event, execution_dat...
{code}

The init container should catch the unavailable DB and retry.

  was:
If one runs:

```
~/git/incubator-airflow/scripts/ci/kubernetes$ kubectl apply 
--namespace=testing-airflow -f kube/
```

The container fails with:

```
sqlalchemy.exc.OperationalError: (psycopg2.OperationalError) could not 
translate host name "postgres-airflow" to address: Name or service not known
```

...and then the pod tries to go into a Running state before 
CrashLoopBackOff'ing about:

```
sqlalchemy.exc.InvalidRequestError: This Session's transaction has been rolled 
back due to a previous exception during flush. To begin a new transaction with 
this Session, first issue Session.rollback(). Original exception was: 
(psycopg2.ProgrammingError) relation "log" does not exist
LINE 1: INSERT INTO log (dttm, dag_id, task_id, event, execution_dat...
```

The init container should catch the unavailable DB and retry.


> Kubernetes init container has no resiliency against unavailable DB
> ------------------------------------------------------------------
>
>                 Key: AIRFLOW-2726
>                 URL: https://issues.apache.org/jira/browse/AIRFLOW-2726
>             Project: Apache Airflow
>          Issue Type: Bug
>            Reporter: Jon Davies
>            Priority: Major
>
> If one runs:
> {code}
> ~/git/incubator-airflow/scripts/ci/kubernetes$ kubectl apply 
> --namespace=testing-airflow -f kube/
> {code}
> The container fails with:
> {code}
> sqlalchemy.exc.OperationalError: (psycopg2.OperationalError) could not 
> translate host name "postgres-airflow" to address: Name or service not known
> {code}
> ...and then the pod tries to go into a Running state before 
> CrashLoopBackOff'ing about:
> {code}
> sqlalchemy.exc.InvalidRequestError: This Session's transaction has been 
> rolled back due to a previous exception during flush. To begin a new 
> transaction with this Session, first issue Session.rollback(). Original 
> exception was: (psycopg2.ProgrammingError) relation "log" does not exist
> LINE 1: INSERT INTO log (dttm, dag_id, task_id, event, execution_dat...
> {code}
> The init container should catch the unavailable DB and retry.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to