Batch Apex for millions of records

Question

I am trying to syc data email from contact to account for 20 million records.Here is the batch i have written to do that.when i run this batch i am getting an error "First error: Aggregate query has too many rows for direct assignment, use FOR loop".Moreover, I am not sure whether this code can handle 20 Million records.Your review and help is much appreciated.

global class EmailSycBatch implements Database.Batchable<SObject> {
    global Database.QueryLocator start(Database.BatchableContext BC) {
        return Database.getQueryLocator(SELECT Name,Id,Primary_Email__c,(SELECT Email,Id,Primary_Indicator__c  FROM Contacts where Primary_Indicator__c=true) FROM Account ); 
    }
    global void execute(Database.BatchableContext BC, List<SObject> scope) {
        for(Account a : scope){
           List<Contact> cc = a.getSObjects('Contacts'); //grab all the related contact records
                if (cc != null) { //if there are related contact records
                for (Contact con: cc) { 
                  if(a.Primary_Email__c != con.Email)
                  a.Primary_Email__c = con.Email;
                }  
            }
         }
     try
        {
            update scope;
        }
        catch(Exception ex)
        {
            throw new TriggerException('Error in batch UpdateAccountBatch on method Batchable execute with the following error message'+ex.getMessage() + 'Stack Trace = ' +ex.getStackTraceString() );
        }   
    }
    global void finish(Database.BatchableContext info) {
        AsyncApexJob a = [SELECT CreatedBy.Email, CreatedBy.Name, CreatedDate, CompletedDate, Id, JobItemsProcessed, NumberOfErrors, Status, TotalJobItems FROM AsyncApexJob WHERE Id =: info.getJobId()];  
        Messaging.SingleEmailMessage mail = new Messaging.SingleEmailMessage(); 
        String[] toAddresses = new String[] {'XXXXXXXX'}; 
        mail.setToAddresses(toAddresses); //to
        mail.setSubject('Apex batch_field_aggregation:'+a.Status); 
        mail.setPlainTextBody('The batch Apex job created by '+a.CreatedBy.Name+' ('+a.CreatedBy.Email+') processed '+a.TotalJobItems+' batches with '+a.NumberOfErrors+' failures. The process began at '+a.CreatedDate+' and finished at '+a.CompletedDate+'.'); 
        Messaging.sendEmail(new Messaging.SingleEmailMessage[] { mail }); //send
    }
}

Answer 1

Query just the parent records in the initial querylocator, then select additional fields as you need in your execute function:

global Database.QueryLocator start(Database.BatchableContext BC) {
    return Database.getQueryLocator([SELECT Id FROM Account]);
}
global void execute(Database.BatchableContext BC, List<SObject> scope) {
    Map<Id, Account> accounts = new Map<Id, Account>(scope);
    for(Account a :[SELECT Primary_Email__c, (SELECT Email FROM Contacts WHERE Primary_Indicator__c=true) WHERE Id IN :scope]){
        for(Contact con: a.Contacts) {
            if(a.Primary_Email__c != con.Email) {
                accounts.get(record.Id).Primary_Email__c = con.Email;
            }  
        }
    }

A few other notes: a sub-query from a query is never null. Also, don't query fields that you don't need. You'll see how I reduced the complexity to just two fields, since that's all I need to know to perform the update logic. Also, you get Id "for free", so you never need to include it in a query.

Additional notes: You're okay with the query as long as it's under 50,000,000 rows. After that, you'll have to figure out how to break up your batches.