Managing Data Growth and Improving Performance

MWARE ESB when in use, will store metadata and runtime data in its connected databases. For example, it stores APIs, applications, subscriptions, and tokens that are created by users. Metadata related to applications and APIs are not been written to the databases frequently. However, as runtime data depends on different attributes such as the number of users, number of connected applications, and usage patterns, having a considerable load on the system will result in runtime data accumulating slowly over time. This will result in high data growth of the tables and this in return will negatively impact the performance of the system.

Invalid access tokens, revoked access tokens, Registry transaction-related logs, authorization codes, and user sessions are some of the runtime data that gets stored in these databases. When you carefully analyze this data, you will see that you do not always need to keep this data, other than for audit purposes. Therefore, this data could be cleaned up periodically.

MWARE ESB provides two methods to do the cleanup.

Regular cleanup is good for regular housekeeping which will slow down unused token growth. However, during peak windows, regular cleaning may be less effective due to potential query slowdowns.

Warning

The deep cleaning removes unused tokens, sessions, and registry data, and prevent the tables from continuously growing, which can improve performance in the production servers. Choose the appropriate cleanup method based on the server's traffic level.

Regular Cleaning

This cleanup is done within the product. It cleans up unused token related data during the runtime. This is an event-based cleaning process that involves the cleaning of specific entries based on specific user actions. For example, when an access token is revoked, the token will be picked from the IDN_OAUTH2_ACCESS_TOKEN table and moved into the IDN_OAUTH2_ACCESS_TOKEN_AUDIT table to be audited. In addition to revoked tokens, inactive and expired tokens also accumulate in this table. This table is not used by MWARE ESB. These tokens are kept in the database for logging and audit purposes, but they can have a negative impact on the server's performance over time. Therefore, it is recommended to clean them.

Note

From MWARE ESB 2.6.0 onwards it is configured by default to trigger token clean up during token generation, token refreshing, and token revocation. Therefore, when the state of the token (TOKEN_STATE) is changed during any of the latter mentioned processes for tokens that were in the ACTIVE state before, by default, such tokens will be removed from the IDN_OAUTH2_ACCESS_TOKEN table and stored in an audit table. Therefore, you don't need to manually clean up the unused tokens as guided below from API-M 2.6.0 onwards. However, you need to configure MWARE ESB to perform regular cleaning.

Configuring MWARE ESB to perform regular cleaning

MWARE ESB triggers token cleanup during the following instances.

  • Token generation
  • Token refresh
  • Token revocation

Note

The regular cleanup procedure is enabled by default, and it cleans up the Inactive, Revoked, and Expired tokens. If you disabled this procedure and after some time you want to enable this cleanup procedure, it is better to clean the access token table using the deep cleaning process and thereafter enable the feature.

To enable or disable token cleanup, open the <API-M_HOME>/repository/conf/deployment.toml file and do the following changes. (Add the configuration if it does not exist in the deployment.toml file)

[oauth.token_cleanup]
enable = true
retain_access_tokens_for_auditing = true
Property Description
enable Set this property to true to enable token cleanup.
Set this property to false to disable token cleanup.
retain_access_tokens_for_auditing Set this property to true to move the old, invalid tokens to the Audit table when token cleaning is enabled.
Set this property to false if you do not wish to store old tokens in the Audit table.

Deep Cleaning

In this cleaning method, all the remaining token data, session data, and Registry data can be cleaned up using separate stored procedures for each type of data. The unused data is periodically analyzed and removed through a stored procedure that runs against the database. In the deep cleaning process, the validity of the data is checked in every record. If unused or old data is detected, the stored procedure will clean them. There are three stored procedures provided that could be used to do the following three cleanups.

  • Token cleanup
  • Session cleanup
  • Registry cleanup

Enabling deep cleaning (token, session, and Registry cleanup)

This will remove the old and invalid tokens, sessions, and Auth codes, which cannot be cleaned by the products inbuilt cleanup process.

Tip

It is safe to run these steps in read-only mode or during a time when traffic on the server is low. However, the latter is not mandatory if you are evaluating or testing the product. However, if you are running a production deployment with a high volume of traffic, it is recommended that you run this periodically as the unused tokens, sessions count, etc. can be high.

  1. Take a backup of the running database.

  2. Optionally, set up the database dump in a test environment and test it for any issues.

    For more information on setting up a database dump, go to the MySQL, SQL Server, and Oracle official documentation.

    Tip

    MWARE recommends testing the cleanup scripts and stored procedures before running or configuring them against your production database server.

  3. Select the appropriate cleanup script, based on your database, from here, and run it on the production database.

    This takes a backup of the necessary tables, turns off SQL updates, and cleans the database of unused tokens. However, when running the script on production make sure the script has been tested at least once in a lower environment.

    Select the token-cleanup script to clean up the tokens, the sessiondata-cleanup script to cleanup the session data, and the registry-cleanup script to clean up the unused data in the Registry.

  4. After the cleanup is over, start the MWARE ESB and test it thoroughly for any issues.

    You can also schedule a cleanup task that will automatically run after a given period of time as shown in the examples below:

    Schedule a cleanup task for MySQL

    USE 'WSO2AM_DB';
DROP EVENT IF EXISTS 'cleanup_tokens_event';
CREATE EVENT 'cleanup_tokens_event'
ON SCHEDULE
    EVERY 1 WEEK STARTS '2018-01-01 00:00.00'
DO
    CALL 'WSO2AM_DB'.'cleanup_tokens'();
-- 'Turn on the event_scheduler'
SET GLOBAL event_scheduler = ON;

    Schedule a cleanup task for SQL Server

    USE WSO2AM_DB ;
GO
-- Creates a schedule named CleanupTask.  
-- Jobs that use this schedule execute every day when the time on the server is 01:00.  
EXEC sp_add_schedule
@schedule_name = N'CleanupTask' ,
@freq_type = 4,
@freq_interval = 1,
@active_start_time = 010000 ;
GO
-- attaches the schedule to the job BackupDatabase
EXEC sp_attach_schedule
@job_name = N'BackupDatabase',
@schedule_name = N'CleanupTask' ;
GO

Replace WSO2AM_DB with the name of your MWARE ESB database in the above script.

Top