Skip to content

Latest commit

 

History

History
109 lines (72 loc) · 7.32 KB

Kafka Vertica Scheduler Upgrade.md

File metadata and controls

109 lines (72 loc) · 7.32 KB
Raw

Upgrade AMWS 1.4 (or earlier) for Single Kafka Vertica Scheduler Usage

The CAF Audit Management Web Service (AMWS) has been modified to use a single Kafka-Vertica scheduler for all tenants. See CAF-1377 for further details. The purpose of this document is to provide a set of upgrade instructions for both the AMWS as well as any existing Audit scheduler entries in Vertica. The upgrade steps are relevant to AMWS version 1.4 or earlier.

Some of the upgrade steps require the use of a database tool that can connect to the Vertica database to run SQL and drop schema objects. It has been assumed that DbVisualizer Free For Vertica will be used for these steps.

Upgrade Steps

  1. Make a record of the Vertica service and loader account information as specified in the AMWS configuration.

  2. Destroy the existing AMWS and any Kafka-Vertica Audit scheduler applications in Marathon. Existing scheduler applications will be registered under the caf-audit-schedulers group in Marathon. The caf-audit-schedulers group can then be removed as well.

  3. Create a single Kafka-Vertica Audit scheduler to handle audit events for all tenants.

    3.1 Create a new Kafka-Vertica scheduler

    The vkconfig script, which comes pre-packaged and installed with the Vertica rpm, should be used with the scheduler command and --add option to do this. Note that the database loader account information recorded in step 1 is required here:

     	sudo /opt/vertica/packages/kafka/bin/vkconfig scheduler --add 
 		--config-schema auditscheduler 
 		--brokers [BROKERS] 
 		--username "[LOADER-USERNAME]" 
 		--password [LOADER-PASSWORD] 
 		--operator "\"[OPERATOR]\""
 		--frame-duration [FRAME-DURATION]

    where:

    • [BROKERS] - This specifies the kafka broker(s) to be used, it is formatted as a comma separated list of address:port endpoints.
    • [LOADER-USERNAME] - This is the vertica database loader account name recorded in step 1 (e.g. caf-audit-loader).
    • [LOADER-PASSWORD] - This is the password for the vertica database loader account recorded in step 1.
    • [OPERATOR] - This is the vertica database loader account name recorded in step 1 (e.g. caf-audit-loader).
    • [FRAME-DURATION] - Configure in accordance with expected throughput and acceptable latency.

    Example:

     	sudo /opt/vertica/packages/kafka/bin/vkconfig scheduler --add 
 		--config-schema auditscheduler 
 		--brokers <kafka broker>:<port number> 
 		--username "caf-audit-loader" 
 		--password c@FaL0Ad3r 
 		--operator "\"caf-audit-loader\""
 		--frame-duration 00:05:00

    Note: This step assumes you have followed the pre-requisite steps after Vertica installation for AMWS usage. See here for further details which describes the database role and service accounts to be created.

    3.2 Grant user access to the new Kafka-Vertica scheduler

    You need to grant the AMWS database user access to objects contained within the new Kafka-Vertica scheduler schema. This step is required in order to support AMWS calls to keep Vertica topic configuration and Kafka topic partitions consistent. Run the following SQL against the Vertica database:

     	GRANT USAGE ON SCHEMA auditscheduler TO "[SERVICE-USERNAME]";

    where:

    • [SERVICE-USERNAME] - This is the vertica database service account name recorded in step 1 (e.g. caf-audit-service).

  4. Deploy the AMWS and Kafka-Vertica Audit scheduler using the Chateau deployment tool set.

     ./deploy-service.sh audit

  5. Perform database clean-up.

    In order for all tenants to be managed by the new Kafka-Vertica scheduler, existing audit scheduler and tenant related rows need to be removed from the Vertica database. Tenants can then be registered via the AMWS afterwards.

    5.1 Make a record of all tenants in the system

    Make a record of all tenant identifiers registered in the system. Run the following SQL against the Vertica database to do this:

     SELECT REPLACE(schema_name,'account_','') as TenantId 
 FROM schemata 
 WHERE schema_name LIKE 'account_%';
    5.2 Clean-up database rows.

    Existing scheduler and tenant related entries in the Vertica database need cleaned-up before tenant registration can be performed. This clean-up can be done using the DbVisualizer database tool.

    ####### 5.2.1 Audit Scheduler Rows. Existing audit scheduler schemas can be identified by the auditscheduler_ prefix followed by the tenantId.

    For each of the tenants that has a schema with the auditscheduler_ prefix, use DbVisualizer to remove each schema by right clicking on the schema name under Schemas in the object tree and select the Drop Schema... option. Make sure you check Cascade Objects on the resulting dialog before clicking on Execute. Then click Yes to confirm.

    ####### 5.2.2 Tenant Rows. Existing tenant related schemas can be identified by the account_ prefix followed by the tenantId.

    ####### 5.2.2.1 Tenant Tables. Each tenant related schema will comprise a number of tables. These include one or more application specific audit tables for storing audit event information as well as a table for storing rejection events. The application audit tables are identified by a Audit prefix and appended with the application name. The rejection table is named kafka_rej.

    For each of the tenants recorded in step 5.1, both their application audit and kafka rejection tables need removed. Using DbVisualizer, right click on the table under the tenant related schema and select the Drop Table... option. Click on Execute and then Yes to confirm.

    ####### 5.2.2.2 Tenant per Application Rows. Tenant per application rows also need removed as part of the clean-up task for all tenants recorded in step 5.1. Using DbVisualizer, right click on the TenantApplications table under the AuditManagement schema in the object tree and select the Empty Table... option. Then click Execute on the resulting dialog followed by Yes to confirm.

  6. Register tenants.

    Now that the database clean-up is complete, all tenants can be registered using the AMWS. For each tenant recorded in step 5.1, issue the following cURL command:

     curl -X POST --header 'Content-Type: application/json' --header 'Accept: application/json' -d '{
   "tenantId": "[TENANTID]",
   "application": [
     "[APPLICATION]"
   ]
 }' 'http://[HOSTNAME]:[SERVICEPORT]/caf-audit-management/v1/tenants'

    where:

    • [TENANTID] - This identifies the tenant identifier as recorded in step 5.1.
    • [APPLICATION] - This identifies the application name to use for tenant registration.
    • [HOSTNAME] - This identifies the host machine.
    • [SERVICEPORT] - This property specifies the external port number on the host machine that will be forwarded to the containers internal 8080 port. This port is used to call the CAF Audit Management Service web service.

    Note that this steps requires the cURL executable.

    To verify, make sure a new row has been added to the tenantApplications table under the AuditManagement schema in the object tree for the specified tenant identifier and application. Further, additional application specific audit and kafka rejection tables will be created under the existing tenant related schema for the specified tenant.