Ignoring Payload table from validation (mysql-time-machine#56)

* Adding support for validator in Replicator v016

* Formatting updates ValidationService

* Validator silently set to null if broker not in config

* Removing pungent code

* refactoring augmenter context & bug fixes

* Upgrading the LOG.warn to a RuntimeException in order to die out as soon as a QUERY_EVENT is encountered which does not have an eventTable

Additionally modified the shouldProcess condition to skip over events related to any schema other than the replicatedSchema (the chosen one)

* Because apparently doing (String) var == (String) var2 (where var and var2 have the same string value) IS FALSE

BUT

If you do (String) var.equals( (String) var2 ) everything works. &shrug; I can't explain it, and apparently that isn't how it's *supposed* to work, but if you look at previous CI builds you see it fails.

* Integration tests for Replicator validator

* Formatting updates

* Fixing kafka message consumption

* Removing code smells for sonarcloud.io

* Added unit test to improve coverage

* Adding more information to the 'Severe environment error', as this error can also indicate a missing table in the ActiveSchema rather than just binlog_row_image=MINIMAL

* Added support for a new configuration variable to enable the ActiveSchema management to fall back to the replicant schema in the event of a missing table.

This is useful in a very narrow set of circumstances, as generally you are much better off properly bootstrapping and starting from a known good state. But, occasionally reality doesn't allow for this, thus this explicit option.

* Unit test updates to improve coverage

* Improving conditional coverage

* Further improving unit test coverage

* Triggering travis build

* Untested patch: for some reason my integration tests fail on my local pc. This patch attempts to revert my fallback behaviour to match the previous code (pre RC2).

* I thought I was being smart by replacing the previous use of SHOW TABLES LIKE ? with the already created SHOW_CREATE_TABLE_SQL var, turns out I overlooked a very important aspect of SHOW TABLES LIKE:
	graceful failure

SHOW CREATE TABLE will throw a fit if the table doesn't actually exist, whereas SHOW TABLES LIKE will gladly return you an empty resultset.

* Providing a default value for this configuration option, as without it we throw an NullPointerException in the event that it isn't provided in the config file.

* Validation task moved to logger debug

* Added field to ignore

* Setting default back to HBASE for integration tests

* Removing code smell

* Adding support for validator in Replicator v016 (mysql-time-machine#51)

* Adding support for validator in Replicator v016

* Formatting updates ValidationService

* Validator silently set to null if broker not in config

* Removing pungent code

* Integration tests for Replicator validator

* Formatting updates

* Fixing kafka message consumption

* Removing code smells for sonarcloud.io

* Added unit test to improve coverage

* Unit test updates to improve coverage

* Improving conditional coverage

* Further improving unit test coverage

* Triggering travis build

* Adding an ignore-field for validator (mysql-time-machine#52)

* Adding support for validator in Replicator v016

* Formatting updates ValidationService

* Validator silently set to null if broker not in config

* Removing pungent code

* Integration tests for Replicator validator

* Formatting updates

* Fixing kafka message consumption

* Removing code smells for sonarcloud.io

* Added unit test to improve coverage

* Unit test updates to improve coverage

* Improving conditional coverage

* Further improving unit test coverage

* Triggering travis build

* Validation task moved to logger debug

* Added field to ignore

* Setting default back to HBASE for integration tests

* Removing code smell

* Sending shard and original table info downstream

* Mutation generator code cleanup

* Fixing tests to incorporate shard and table forward

* Avoiding null metrics config

* Null check for metrics

* Removing code smell

* Ignoring Payload table name from validation

* Moved conditional to a new line

Co-authored-by: Bosko Devetak <bosko.devetak@gmail.com>
Co-authored-by: aka-netzero <steve.aurigema@gmail.com>

Author: 3 people

mysql-replicator-applier/src/main/java/com/booking/replication/applier/hbase/writer/HBaseTimeMachineWriter.java

@@ -43,6 +43,7 @@ public class HBaseTimeMachineWriter implements HBaseApplierWriter {
     HBaseApplierMutationGenerator mutationGenerator;
     private final boolean dryRun;
     private final ValidationService validationService;
+    private final String payloadTableName;
 
     Connection connection;
     Admin admin;
@@ -57,6 +58,8 @@ public HBaseTimeMachineWriter(Configuration hbaseConfig,
 
         this.validationService = validationService;
 
+        this.payloadTableName = (String) configuration.get(HBaseApplier.Configuration.PAYLOAD_TABLE_NAME);
+
         this.metrics = Metrics.getInstance(configuration);
 
         this.hbaseSchemaManager = hbaseSchemaManager;
@@ -256,7 +259,9 @@ private void writeToHBase(Long threadID) throws IOException {
 //            table.close();
 
             for (HBaseApplierMutationGenerator.PutMutation mutation : tableMutations){
-                if (validationService != null) validationService.registerValidationTask(mutation.getTransactionUUID(), mutation.getSourceRowUri(), mutation.getTargetRowUri());
+                if (validationService != null && !tableName.equalsIgnoreCase(payloadTableName)) {
+                    validationService.registerValidationTask(mutation.getTransactionUUID(), mutation.getSourceRowUri(), mutation.getTargetRowUri());
+                }
             }
 
             long timeEnd = System.currentTimeMillis();