ADB Control known issues

Daria Barysheva
Issue From Fixed Workaround

In AltLinux 8, the Docker engine update may break the docker-compose package, which causes the ADB Control start to fail

 — 

AltLinux fix

To fix /usr/lib/python3/site-packages/compose/service.py, replace the following:

container.image_config['ContainerConfig'].get('Volumes') or {}

by:

container.image_config['Config'].get('Volumes') or {}

After the ADB Control service is configured to use LDAP authentication with FreeIPA type, a list of groups is not loaded when you create a new mapping on the Users → LDAP group mapping page

4.0.0

 — 

Use basic authentication or LDAP with MSAD type (Microsoft Active Directory)

When a cluster is upgraded from 6.27.1.56 to 6.27.1.58, an error occurs if external ADB Control is used

4.10.3

4.10.3

Upgrade a cluster with ADB Control before upgrading clusters that use this ADB Control as external (in External ADB Control server parameters)

After the transaction cancellation, plan nodes in the Plan & progress section will have the executing status and display the increasing execution time

4.8.8

4.9.1

Keep in mind that cancelled queries do not have executing processes

ADB Control 4.8.8 implements changes in API that affect communication between the main and external ADB Control services of different versions.

When an ADB cluster with the main ADB Control is upgraded to 6.27.1.56.b1, clusters of older versions that use this ADB Control as external will not be able to perform Reconfigure/Start actions with the ADB Control service

4.8.8

4.8.8

In order to make all ADB Control service actions to work, upgrade clusters with the external ADB Control and versions older than 6.27.1.56.b1 to 6.27.1.56.b1

After ADB Control upgrade to 4.3.2, the Export job settings (that are defined on the Configuration → Job policy → Metrics offload page) are reset to defaults

4.3.2

4.3.2

Fill in and save the required settings. The problem does not affect future upgrades

Standby activation after switching from master fails on clusters with ADB Control/ADBM installed

4.3.1

4.3.2

Use external ADB Control/ADBM

In ADB Control 4.2.1, you cannot register new external ADB clusters with versions earlier than 6.23.3

4.2.1

4.2.1

Update ADB cluster to 6.23.3

Switching off sending metrics of inner queries to ADB Control may cause hangs of ADB queries

4.1.0

4.10.3

Turn on the GUC that enables sending metrics of inner queries in ADB Control or in ADB (as follows):

$ gpconfig -c adcc.monitor_inner_queries -v on
$ gpstop -u

Commands appear in the Running status on the Monitoring → Commands → Online page but after 14-16 seconds they freeze in the Unknown status (with no metrics updated). After the command completion, all metrics become available on the Monitoring → Commands → History page, but the Unknown status persists

4.1.0

4.2.0

  1. Using the psql or other SQL IDE, set the adcc user password via the following command:

    ALTER USER adcc WITH ENCRYPTED PASSWORD '<password>';

  2. Open the Configuration → Clusters page in the ADCC UI and go to editing the cluster connection.

  3. In the Password field, set the <password> value from the step 1. Click Save.

There is a potential memory leak with specific queries affected by the gpperfmon engine. If you use queries against huge tables with UDF that generate inner per-tuple queries, you may experience the significant RAM consumption, and even get OOM errors. The example is the information_schema.columns view, which can cause memory leak when being selected against a large system catalog

4.0.0

4.1.0

The possible solution is to switch off monitoring metrics (this operation requires cluster restart):

  1. Remove gpadcc from shared_preload_libraries, e.g.:

    $ gpconfig -c shared_preload_libraries -v ""

  2. Switch off the following parameters:

    $ gpconfig -c gp_enable_gpperfmon -v off
$ gpconfig -c adcc.explain_log_verbose -v off
$ gpconfig -c adcc.explain_log_analyze -v off

  3. Restart your cluster:

    $ gpstop -a -M fast
$ gpstart a

Remember that this workaround does not disable the ADCC UI and Arenadata DB Backup Manager

Backend ADCC server cannot connect to ADCC UI server. Data update may not always work on the details page

4.0.0

4.1.0

  1. Connect to the ADCC server via SSH.

  2. Go to the /opt/adcc directory.

  3. Edit the docker-compose.yaml file. For the adcc_ui_server and adcc_backend_server services, change the value of the ADCC_EUREKA_CLIENT_PREFER_IP_ADDRESS environment variable to true.

  4. Run the command and wait until all Docker containers stop:

    $ docker-compose down

  5. Run the following command:

    $ docker-compose up -d

It is impossible to configure the Data cleanup job on the Configuration page. In general, ADCC can work without data cleanup, but you need to monitor the metrics database size

4.0.0

4.1.0

  1. Connect to the ADCC server via SSH.

  2. Go to the /opt/adcc directory.

  3. Edit the docker-compose.yaml file. For the adcc_scheduler service, add the following environment variable: ADCC_SCHEDULER_GRPC_PORT: 6572.

  4. Run the command and wait until all Docker containers stop:

    $ docker-compose down

  5. Run the following command:

    $ docker-compose up -d
Found a mistake? Seleсt text and press Ctrl+Enter to report it