There are numerous GDSCTL CONFIG commands that you can use to obtain the health status of individual shards, shardgroups, shardspaces, and shard directors.

Monitoring a shard is just like monitoring a normal database, and standard Oracle best practices should be used to monitor the individual health of a single shard. However, it is also important to monitor the overall health of the entire sharded environment. The GDSCTL commands can also be scripted and through the use of a scheduler and can be done at regular intervals to help ensure that everything is running smoothly. When using Oracle GoldenGate for replication it is also important to monitor the lag of each replication stream.

Oracle Enterprise Manager Cloud Control lets you discover, monitor, and manage the components of a sharded database.

Sharded database targets are found in the All Targets page.

Figure 10-1 Sharded Databases in the All Targets Refine Search pane

Description of "Figure 10-1 Sharded Databases in the All Targets Refine Search pane"

The target home page for a sharded database shows you a summary of the sharded database components and their statuses.

To monitor sharded database components you must first discover them. See Discovering Sharded Database Components for more information.

Summary

The Summary pane, in the top left of the page, shows the following information:

• Sharded database name

• Sharded database domain name

• Shard catalog name. You can click the name to view more information about the shard catalog.

• Shard catalog database version

• Sharding method used to shard the database

• Replication technology used for high availability

• Number and status of the shard directors

• Master shard director name. You can click the name to view more information about the master shard director.

Figure 10-2 Sharded Database Summary pane

Description of "Figure 10-2 Sharded Database Summary pane"

Shard Load Map

The Shard Load Map, in the upper right of the page, shows a pictorial graph illustrating how transactions are distributed among the shards.

Figure 10-3 Sharded Database Shard Load Map

Description of "Figure 10-3 Sharded Database Shard Load Map"

You can select different View Levels above the graph.

• Database

  The database view aggregates database instances in Oracle RAC cluster databases into a single cell labeled with the Oracle RAC cluster database target name. This enables you to easily compare the total database load in Oracle RAC environments.

• Instance

  The instance view displays all database instances separately, but Oracle RAC instances are grouped together as sub-cells of the Oracle RAC database target. This view is essentially a two-level tree map, where the database level is the primary division, and the instance within the database is the secondary division. This allows load comparison of instances within Oracle RAC databases; for instance, to easily spot load imbalances across instances.

• Pluggable Database

  Although the PDB option is shown, PDB is not supported for Oracle Sharding in the current release.

Notice that the cells of the graph are not identical in size. Each cell corresponds to a shard target, either an instance or a cluster database. The cell size (its area) is proportional to the target database's load measured in average active sessions, so that targets with a higher load have larger cell sizes. Cells are ordered by size from left to right and top to bottom. Therefore, the target with the highest load always appears as the upper leftmost cell in the graph.

You can hover your mouse pointer over a particular cell of the graph to view the total active load (I/O to CPU ration), CPU, I/O, and wait times. Segments of the graph are colored to indicate the dominant load:

• Green indicates that CPU time dominates the load

• Blue indicates that I/O dominates the load

• Yellow indicates that WAIT dominates the load

Members

The Members pane, in the lower left of the page, shows some relevant information about each of the components.

Figure 10-4 Sharded Database Members pane

Description of "Figure 10-4 Sharded Database Members pane"

The pane is divided into tabs for each component: Shardspaces, Shardgroups, Shard Directors, and Shards. Click on a tab to view the information about each type of component

• Shardspaces

  The Shardspaces tab displays the shardspace names, status, number of chunks, and Data Guard protection mode. The shardspace names can be clicked to reveal more details about the selected shardspace.

• Shardgroups

  The Shardgroups tab displays the shardgroup names, status, the shardspace to which it belongs, the number of chunks, Data Guard role, and the region to which it belongs. You can click the shardgroup and shardspace names to reveal more details about the selected component.

• Shard Directors

  The Shard Directors tab displays the shard director names, status, region, host, and Oracle home. You can click the shard director names can be clicked to reveal more details about the selected shard director.

• Shards

  The Shards tab displays the shard names, deploy status, status, the shardspaces and shardgroups to which they belong, Data Guard roles, and the regions to which they belong. In the Names column, you can expand the Primary shards to display the information about its corresponding Standby shard. You can hover the mouse over the Deployed column icon and the deployment status details are displayed. You can click on the shard, shardspace, and shardgroup names to reveal more details about the selected component.

Services

The Services pane, in the lower right of the page, shows the names, status, and Data Guard role of the sharded database services. Above the list is shown the total number of services and an icon showing how many services are in a particular status. You can hover your mouse pointer over the icon to read a description of the status icon.

Figure 10-5 Sharded Database Services pane

Description of "Figure 10-5 Sharded Database Services pane"

Target Navigation

The Target Navigation pane gives you easy access to more details about any of the components in the sharded database.

Description of the illustration shard_em_navicon.png

Clicking the navigation tree icon on the upper left corner of the page opens the Target Navigation pane. This pane shows all of the discovered components in the sharded database in tree form.

Expanding a shardspace reveals the shardgroups in them. Expanding a shardgroup reveals the shards in that shardgroup.

Any of the component names can be clicked to view more details about them.

• Discovering Sharded Database Components
  In Enterprise Manager Cloud Control, you can discover the shard catalog and shard databases, then add the shard directors, sharded databases, shardspaces, and shardgroups using guided discovery.

Use the SHARDS() clause to query Oracle-supplied tables to gather performance, diagnostic, and audit data from V$ views and DBA_* views.

The shard catalog database can be used as the entry point for centralized diagnostic operations using the SQL SHARDS() clause. The SHARDS() clause allows you to query the same Oracle supplied objects, such as V$, DBA/USER/ALL views and dictionary objects and tables, on all of the shards and return the aggregated results.

As shown in the examples below, an object in the FROM part of the SELECT statement is wrapped in the SHARDS() clause to specify that this is not a query to local object, but to objects on all shards in the sharded database configuration. A virtual column called SHARD_ID is automatically added to a SHARDS()-wrapped object during execution of a multi-shard query to indicate the source of every row in the result. The same column can be used in predicate for pruning the query.

A query with the SHARDS() clause can only be run on the shard catalog database.

SQL> SELECT shard_id, callspersec FROM SHARDS(v$servicemetric)
 WHERE service_name LIKE 'oltp%' AND group_id = 10;

SQL> SELECT table_name, partition_name, blocks, num_rows
 FROM SHARDS(dba_tab_partition) p
 WHERE p.table_owner= :1;

SQL> select ORA_SHARD_ID, INSTANCE_NAME from SHARDS(sys.v_$instance);

    ORA_SHARD_ID INSTANCE_NAME
    ------------ ----------------
               1 sh1
              11 sh2
              21 sh3
              31 sh4

SQL> select ORA_SHARD_ID, INSTANCE_NAME
 from SHARDS(sys.v_$instance)
 where ORA_SHARD_ID=21;

    ORA_SHARD_ID INSTANCE_NAME
    ------------ ----------------
              21 sh3

New shards can be added to an existing sharded database environment to scale out and to improve fault tolerance.

For fault tolerance, it is beneficial to have many smaller shards than a few very large ones. As an application matures and the amount of data increases, you can add an entire shard or multiple shards to the SDB to increase capacity. 

When you add a shard to a sharded database, if the environment is sharded by consistent hash, then chunks from existing shards are automatically moved to the new shard to rebalance the sharded environment.

When using user-defined sharding, populating a new shard with data may require manually moving chunks from existing shards to the new shard using the GDSCTL split chunk and move chunk commands.

Oracle Enterprise Manager Cloud Control can be used to help identify chunks that would be good candidates to move, or split and move to the new shard.

When you add a shard to the environment, verify that the standby server is ready, and after the new shard is in place take backups of any shards that have been involved in a move chunk operation.

The process of redistributing data between shards, triggered by a change in the number of shards, is called resharding. Automatic resharding is a feature of the system-managed sharding method that provides elastic scalability of an SDB.

Sometimes data in an SDB needs to be migrated from one shard to another. Data migration across shards is required in the following cases:

• When one or multiple shards are added to or removed from an SDB

• When there is skew in the data or workload distribution across shards

The unit of data migration between shards is the chunk. Migrating data in chunks guaranties that related data from different sharded tables are moved together.

When a shard is added to or removed from an SDB, multiple chunks are migrated to maintain a balanced distribution of chunks and workload across shards.

Depending on the sharding method, resharding happens automatically (system-managed) or is directed by the user (composite). The following figure shows the stages of automatic resharding when a shard is added to an SDB with three shards.

Figure 10-6 Resharding an SDB

Description of "Figure 10-6 Resharding an SDB"

A particular chunk can also be moved from one shard to another, when data or workload skew occurs, without any change in the number of shards. In this case, chunk migration can be initiated by the database administrator to eliminate the hot spot.

RMAN Incremental Backup, Transportable Tablespace, and Oracle Notification Service technologies are used to minimize impact of chunk migration on application availability. A chunk is kept online during chunk migration. There is a short period of time (a few seconds) when data stored in the chunk is available for read-only access only.

FAN-enabled clients receive a notification when a chunk is about to become read-only in the source shard, and again when the chunk is fully available in the destination shard on completion of chunk migration. When clients receive the chunk read-only event, they can either repeat connection attempts until the chunk migration is completed, or access the read-only chunk in the source chunk. In the latter case, an attempt to write to the chunk will result in a run-time error.

It may become necessary to remove a shard from the sharded database environment, either temporarily or permanently, without losing any data that resides on that shard.

For example, removing a shard might become necessary if a sharded environment is scaled down after a busy holiday, or to replace a server or infrastructure within the data center. Prior to decommissioning the shard, you must move all of the chunks from the shard to other shards that will remain online. As you move them, try to maintain a balance of data and activity across all of the shards.

If the shard is only temporarily removed, keep track of the chunks moved to each shard so that they can be easily identified and moved back once the maintenance is complete.

You can add Data Guard standby shards to an Oracle Sharding environment; however there are some limitations.

1. Stop all global services on the shard using the GDSCTL command STOP SERVICE.
2. Disable all global services on the shard using the GDSCTL command DISABLE SERVICE.
3. Convert the shard to a snapshot standby using the procedure described in the Data Guard documentation.

   At this point, the shard remains part of the sharded database, but will not accept connections which use the sharding key.

You can manage database shards using Oracle Enterprise Manager Cloud Control

To manage shards using Cloud Control, they must first be discovered. Because each database shard is a database itself, you can use standard Cloud Control database discovery procedures.

The following topics describe shard management using Oracle Enterprise Manager Cloud Control:

• Validating a Shard
  Validate a shard prior to adding it to your Oracle Sharding deployment.
• Adding Primary Shards
  Use Oracle Enterprise Manager Cloud Control to add a primary shards to your Oracle Sharding deployment.
• Adding Standby Shards
  Use Oracle Enterprise Manager Cloud Control to add a standby shards to your Oracle Sharding deployment.
• Deploying Shards
  Use Oracle Enterprise Manager Cloud Control to deploy shards that have been added to your Oracle Sharding environment.

You can manage shards in your Oracle Sharding deployment using the GDSCTL command-line utility.

The following topics describe shard management using GDSCTL:

• Validating a Shard
  Before adding a newly created shard to a sharding configuration, you must validate that the shard has been configured correctly for the sharding environment.
• Adding Shards to a System-Managed SDB
  Adding shards to a system-managed SDB elastically scales the SDB. In a system-managed SDB chunks are automatically rebalanced after the new shards are added.
• Replacing a Shard
  If a shard fails and is unrecoverable, or if you just want to move a shard to a new host for other reasons, you can replace it using the ADD SHARD -REPLACE command in GDSCTL.

Sometimes it becomes necessary to move a chunk from one shard to another. To maintain scalability of the sharded environment, it is important to attempt to maintain an equal distribution of the load and activity across all shards.

As the environment matures in a composite SDB, some shards may become more active and have more data than other shards. In order to keep a balance within the environment you must move chunks from more active servers to less active servers. There are other reasons for moving chunks:

• When a shard becomes more active than other shards, you can move a chunk to a less active shard to help redistribute the load evenly across the environment.

• When using range, list, or composite sharding, and you are adding a shard to a shardgroup.

• When using range, list, or composite sharding, and you a removing a shard from a shardgroup.

• After splitting a chunk it is often advisable to move one of the resulting chunks to a new shard.

When moving shards to maintain scalability, the ideal targets of the chunks are shards that are less active, or have a smaller portion of data. Oracle Enterprise Manager and AWR reports can help you identify the distribution of activity across the shards, and help identify shards that are good candidates for chunk movement.

You can move chunks from one shard to another in your Oracle Sharding deployment using Oracle Enterprise Manager Cloud Control.

1. From a shardspace management page, open the Shardspace menu, located in the top left corner of the Sharded Database target page, and choose Manage Shardgroups.
2. Select a shardgroup in the list and click Move Chunks.
3. In the Move Chunks dialog, select the source and destination shards between which to move the chunks.
4. Select the chunks that you want to move by choosing one of the options.

   • Enter ID List: enter a comma separates list of chunk ID numbers

   • Select IDs From Table: click the chunk IDs in the table

5. Indicate when the chunk move should occur.

   • Immediately: the chunk move is provisioned upon confirmation

   • Later: schedule the timing of the chunk move using the calendar tool in the adjacent field

6. Click OK.
7. Click the link in the Information box at the top of the page to view the provisioning status of the chunk move.

Splitting a chunk in a sharded database is required when chunks become too big, or only part of a chunk must be migrated to another shard.

Oracle Sharding supports the online split of a chunk. Theoretically it is possible to have a single chunk for each shard and split it every time data migration is required. However, even though a chunk split does not affect data availability, the split is a time-consuming and CPU-intensive operation because it scans all of the rows of the partition being split, and then inserts them one by one into the new partitions. For composite sharding, it is time consuming and may require downtime to redefine new values for the shard key or super shard key.

Therefore, it is recommended that you pre-create multiple chunks on each shard and split them either when the number of chunks is not big enough for balanced redistribution of data during re-sharding, or a particular chunk has become a hot spot.

Even with system-managed sharding, a single chunk may grow larger than other chunks or may become more active. In this case, splitting that chunk and allowing automatic resharding to move one of the resulting chunks to another shard maintains a more equal balanced distribution of data and activity across the environment.

Oracle Enterprise Manager heat maps show which chunks are more active than other chunks. Using this feature will help identify which chunks could be split, and one of the resulting chunks could then be moved to another shard to help rebalance the environment.

You can split chunks in your Oracle Sharding deployment using Oracle Enterprise Manager Cloud Control.

1. Open the Sharded Database menu, located in the top left corner of the Sharded Database target page, and choose Shardspaces.
2. If prompted, enter the shard catalog credentials, select the shard director to manage under Shard Director Credentials, select the shard director host credentials, and log in.
3. Select a shardspace in the list and click Split Chunks.
4. Select the chunks that you want to split by choosing one of the options.

   • Enter ID List: enter a comma separate list of chunk ID numbers

   • Select IDs From Table: click the chunk IDs in the table

5. Indicate when the chunk split should occur.

   • Immediately: the chunk split is provisioned upon confirmation

   • Later: schedule the timing of the chunk split using the calendar tool in the adjacent field

6. Click OK.
7. Click the link in the Information box at the top of the page to view the provisioning status of the chunk split.

Use Oracle Enterprise Manager Cloud Control to create and add a shard director to your Oracle Sharding deployment.

1. Open the Sharded Database menu, located in the top left corner of the Sharded Database target page, and choose Shard Directors.
2. If prompted, enter the shard catalog credentials, select the shard director to manage under Shard Director Credentials, select the shard director host credentials, and log in.
3. Click Create, or select a shard director from the list and click Create Like.

   Choosing Create opens the Add Shard Director dialog with default configuration values in the fields.

   Choosing Create Like opens the Add Shard Director dialog with configuration values from the selected shard director in the fields. You must select a shard director from the list to enable the Create Like option.

4. Enter the required information in the Add Shard Director dialog, and click OK.

   Note:

   If you do not want the shard director to start running immediately upon creation, you must uncheck the Start Shard Director After Creation checkbox.

5. Click OK on the confirmation dialog.
6. Click the link in the Information box at the top of the page to view the provisioning status of the shard director.

Use Oracle Enterprise Manager Cloud Control to edit a shard director configuration in your Oracle Sharding deployment.

1. Open the Sharded Database menu, located in the top left corner of the Sharded Database target page, and choose Shard Directors.
2. If prompted, enter the shard catalog credentials, select the shard director to manage under Shard Director Credentials, select the shard director host credentials, and log in.
3. Select a shard director from the list and click Edit.

   Note that you cannot edit the shard director name, host, or Oracle home.

4. Edit the fields, enter the GSMCATUSER password, and click OK.
5. Click the link in the Information box at the top of the page to view the provisioning status of the shard director configuration changes.

Use Oracle Enterprise Manager Cloud Control to remove shard directors from your Oracle Sharding deployment.

1. Open the Sharded Database menu, located in the top left corner of the Sharded Database target page, and choose Shard Directors.
2. If prompted, enter the shard catalog credentials, select the shard director to manage under Shard Director Credentials, select the shard director host credentials, and log in.
3. Select a shard director from the list and click Delete.
4. Click the link in the Information box at the top of the page to view the provisioning status of the shard director removal.

Create sharded database regions in your Oracle Sharding deployment using Oracle Enterprise Manager Cloud Control.

1. Open the Sharded Database menu, located in the top left corner of the Sharded Database target page, and choose Regions.
2. If prompted, enter the shard catalog credentials, select the shard director to manage under Shard Director Credentials, select the shard director host credentials, and log in.
3. Click Create.
4. Enter a unique name for the region in the Create Region dialog.
5. Optionally, select a buddy region from among the existing regions.
6. Click OK.
7. Click the link in the Information box at the top of the page to view the provisioning status of the region.

Edit sharded database region configurations in your Oracle Sharding deployment using Oracle Enterprise Manager Cloud Control.

1. Open the Sharded Database menu, located in the top left corner of the Sharded Database target page, and choose Regions.
2. If prompted, enter the shard catalog credentials, select the shard director under Shard Director Credentials, select the shard director host credentials, and log in.
3. Select a region from the list and click Edit.
4. Select or remove a buddy region, and click OK.
5. Click the link in the Information box at the top of the page to view the provisioning status of the region configuration changes.

Remove sharded database regions in your Oracle Sharding deployment using Oracle Enterprise Manager Cloud Control.

1. Open the Sharded Database menu, located in the top left corner of the Sharded Database target page, and choose Regions.
2. If prompted, enter the shard catalog credentials, select the shard director under Shard Director Credentials, select the shard director host credentials, and log in.
3. Select a region from the list and click Delete.
4. Click the link in the Information box at the top of the page to view the provisioning status of the region removal.

Create shardspaces in your composite Oracle Sharding deployment using Oracle Enterprise Manager Cloud Control.

1. Open the Sharded Database menu, located in the top left corner of the Sharded Database target page, and choose Shardspaces.
2. If prompted, enter the shard catalog credentials, select the shard director to manage under Shard Director Credentials, select the shard director host credentials, and log in.
3. Click Create.

   Note:

   This option is disabled in the Shardspaces page for a system-managed sharded database.
4. Enter the values in the fields in the Add Shardspace dialog, and click OK.

   • Name: enter a unique name for the shardspace (required)

   • Chunks: Enter the number of chunks that should be created in the shardspace (default 120)

   • Protection Mode: select the Data Guard protection mode (default Maximum Performance)

5. Click the link in the Information box at the top of the page to view the provisioning status of the shardspace.

Learn to create a new shardspace, add shards to the shardspace, create a tablespace set in the new shardspace, and add a partitionset to the sharded table for the added shardspace. Then verify that the partitions in the tables are created in the newly added shards in the corresponding tablespaces.

1. Create a new shardspace, add shards to the shardspace, and deploy the environment.
   1. Connect to the shard catalog database.

      GDSCTL> connect mysdbadmin/mysdbadmin_password

   2. Add a shardspace and add a shardgroup to the shardspace.

      GDSCTL> add shardspace -chunks 8 -shardspace cust_asia
      GDSCTL> add shardgroup -shardspace cust_asia -shardgroup asia_shgrp1 -deploy_as primary -region region3

   3. Add shards

      GDSCTL> add shard -shardgroup asia_shgrp1 –connect shard_host:TNS_listener_port/shard_database_name –pwd GSMUSER_password
      GDSCTL> add shard asia_shgrp1 –connect shard_host:TNS_listener_port/shard_database_name –pwd GSMUSER_password

   4. Deploy the environment.

      GDSCTL> deploy

   Running DEPLOY ensures that all of the previous DDLs are replayed on the new shards and all of the tables are created. The partition is created in the default SYS_SHARD_TS tablespace.
2. On the shard catalog create the tablespace set for the shardspace and add partitionsets to the sharded root table.
   1. Create the tablespace set.

      SQL> CREATE TABLESPACE SET
        TSP_SET_3 in shardspace cust_asia using template
        (datafile size 100m autoextend on next 10M maxsize
         unlimited extent management
         local segment space management auto );

   2. Add the partitionset.

      SQL> ALTER table customers add PARTITIONSET asia VALUES ('ASIA”') TABLESPACE SET TSP_SET_3 ;

   3. When lobs are present, create the tablespace set for lobs and mention the lob storage information in the add partitionset command.

      SQL> alter table customers add partitionset asia VALUES ('ASIA') tablespace set TSP_SET_3 lob(docn) store as (tablespace set LOBTSP_SET_4)) ;

   4. When the root table contains subpartitions, use the store as clause to specify the tablespace set for the subpartitions.

      SQL> alter table customers add partitionset asia VALUES ('ASIA') tablespace set TSP_SET_3 subpartitions store in(SUB_TSP_SET_1, SUB_TSP_SET_2);

   The ADD PARTITIONSET command ensures that the child tables are moved to the appropriate tablespaces.
3. Verify that the partitions in the new shardspace are moved to the new tablespaces.

   Connect to the new shards and verify that the partitions are created in the new tablespace set.

   SQL> select table_name, partition_name, tablespace_name, read_only from dba_tab_partitions;

Create shardgroups in your Oracle Sharding deployment using Oracle Enterprise Manager Cloud Control.

1. Select a shardspace to which to add the shardgroup.
2. Open the Shardspace menu, located in the top left corner of the shardspace target page, and choose Manage Shardgroups.
3. Click Create.
4. Enter values in the Create Shardgroup dialog, and click OK.
5. Click the link in the Information box at the top of the page to view the provisioning status of the shardgroup.

   For example, with the values entered in the screenshots above, the following command is run:

   GDSCTL Command: ADD SHARDGROUP -SHARDGROUP 'north' -SHARDSPACE 'shardspaceora'
    -REGION 'north' -DEPLOY_AS 'STANDBY'

Create services in your Oracle Sharding deployment using Oracle Enterprise Manager Cloud Control.

1. Open the Sharded Database menu, located in the top left corner of the Sharded Database target page, and choose Services.
2. If prompted, enter the shard catalog credentials, select the shard director to manage under Shard Director Credentials, select the shard director host credentials, and log in.
3. Click Create, or select a service from the list and click Create Like.

   Choosing Create opens the Create Service dialog with default configuration values in the fields.

   Choosing Create Like opens the Create Like Service dialog with configuration values from the selected service in the fields. You must select a service from the list to enable the Create Like option.

4. Enter the required information in the dialog, and click OK.

   Note:

   If you do not want the service to start running immediately upon creation, you must uncheck the Start service on all shards after creation checkbox.

5. Click the link in the Information box at the top of the page to view the provisioning status of the service.