6/28/2017 8:28:12 PM
(#1253)
|
6/28/2017 7:16:38 PM
(#1255)
The current code uses the `SERVICE_PROVIDER` interface to fetch dependencies for the Web Server. Refactored the code to enable parameterized dependency injection from the construction which is the standard way of declaring dependencies. In the process, removed legacy constructor from `AzkabanWebServer` and removed legacy class `AzkabanServletContextListener` which is no longer being used.
Injection sanity tests already exist in the code. Tested solo server and it works as well.
|
|
6/27/2017 7:04:20 PM
to #990
Users currently can login by passing their credentials through the query string in the URI in this method:
curl -X POST http://localhost:8081/?action=login\&username=azkaban\&password=azkaban
This leads to sensitive information being logged in access logs and overall isn't a secure method of logging in.
This PR fails these attempts and returns an appropriate warning to users. It also stops the query string from being logged when these requests are made.
This PR includes testing to replicate this user behavior for future regression testing.
|
6/23/2017 3:13:24 PM
JSON blob (#1250)
submitted user should be deserialized from JSON instead of a constant value.
|
6/22/2017 4:12:52 PM
3.30.0
reverts PR #1225 (commit e1f6d3942b29d917aa2afdd94cd2f7ce51a6f9e3)
This plugin breaks the internal build.
|
6/21/2017 10:06:27 PM
that causing schedule end time is set to a past date whenever SLA changes are made.
It's caused by incompatibility of new code #1110 with old schedule metadata. SLA change triggers reloading updated schedules into memory. But instead of using a default schedule end time, which we set to some day in year 2050, it always mistakenly uses a past date as the expiring time of the schedule.
This fixes the issue by setting appropriate end time of the schedule.
|
6/21/2017 7:52:23 PM
(#1235)
Azkaban currently uses `HiveMetaStoreClient` while communicating with the Hive Metastore ref class: `HadoopSecurityManager_H_2_0`
This client does not automatically retry/reconnect in case of transient failures, such as when a metastore is restarted. The RetryingHiveMetaStoreClient does automatically retry/reconnect in case of errors.
This change switches it to use the RetryingHiveMetaStoreClient (which is what the Hive CliDriver uses internally).
NOTE: This introduces a runtime dependency on `hive-exec` core jar. Currently, this is set as a compile only dependency in Azkaban but it actually requires this jar in the runtime and expects to get it from the `HIVE_HOME` environment variable.
Ref: https://github.com/linkedin/gobblin/issues/1506
|
6/21/2017 2:21:57 PM
SaveAction, which was disabled for prior PR. No other changes made.
|
|