GitHub user victor-wong opened a pull request:
https://github.com/apache/spark/pull/17937
Reload credentials file config when app starts with checkpoint file iâ¦
## What changes were proposed in this pull request?
Currently credentials file configuration is recovered from checkpoint file
when Spark Streaming applicatioin is restarted, which will lead to some
unwanted behaviors, for example:
1. Submit Spark Streaming application using keytab file with checkpoint
enabled in yarn-cluster mode.
> spark-submit --master yarn-cluster --principal xxxx --keytab xxx ...
2. Stop Spark Streaming application;
3. Resubmit this application after a period of time (i.e. one day);
4. Credentials file configuration recover from checkpoint file, so value of
"spark.yarn.credentials.file" points to old staging directory (i.e.
hdfs://xxxx/.sparkStaging/application_xxxx/credentials-xxxx, application_xxxx
is the application id of the previous application which was stopped.)
4. When launching executor, ExecutorDelegationTokenUpdater will update
credentials from credentials file immediately. As credentials file was
generated one day ago (maybe older), it has already expired, so after a period
of time the executor keeps failing.
Some useful logs are shown below :
>2017-04-27,15:08:08,098 INFO
org.apache.spark.executor.CoarseGrainedExecutorBackend: Will periodically
update credentials from: hdfs://xxxx/application_xxxx/credentials-xxxx
>2017-04-27,15:08:12,519 INFO
org.apache.spark.deploy.yarn.ExecutorDelegationTokenUpdater: Reading new
delegation tokens from hdfs://xxxx/application_1xxxx/credentials-xxxx-xx
>2017-04-27,15:08:12,661 INFO
org.apache.spark.deploy.yarn.ExecutorDelegationTokenUpdater: Tokens updated
from credentials file.
...
>2017-04-27,15:08:48,156 WARN org.apache.hadoop.ipc.Client: Exception
encountered while connecting to the server :
org.apache.hadoop.ipc.RemoteException(org.apache.hadoop.security.token.SecretManager$InvalidToken):
token (HDFS_DELEGATION_TOKEN token xxxx for xx) can't be found in cache
## How was this patch tested?
manual tests
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/victor-wong/spark fix-credential-file
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/17937.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #17937
----
commit fac97c69b8087fda62b776384539301df0230ae2
Author: jiasheng.wang <[email protected]>
Date: 2017-05-10T09:35:11Z
Reload credentials file config when app starts with checkpoint file in
cluster mode
----
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]