Recently , I inherited an Org with huge customization , which enqueue 100+ Batch Jobs in few scenario. Now, don’t ask me why ?
I remember, few years back, Salesforce had limit of 5 Apex Batch that can be executed at a time, but we had expectation and demand !!! Salesforce introduced Flex queue and we can have 100 Batch apex waiting to be executed, still we are not happy. After all, human is wanting animal.
After 100 Batch apex, all jobs were failing with error System.AsyncException You have exceeded the limit of 100 Jobs in the flex queue. As I explained my life previously in this post, I had to fix this issue as well.
Right way to fix it was to analyze existing code, perform code review, why do we need even customization etc.. However, time was crucial and I had to do something quickly.
Below framework was used to fix the issue
- Create a Custom Object to hold information about batch
- Use an intermediate Apex class to queue Batch Apex and Queueable Apex. If there is no availability in Flex queue then simply dump it in custom object.
- Use a Scheduler which will run every 10 mins, check Flex queue and submit batches from Custom object if limit available
As you can see in below code, syntax to submit batch Apex is almost similar.
Instead of
ID batchprocessid = Database.executeBatch(batchApexInstance,200);
we would be using
ID batchprocessid = AsyncApexFramework.submitBatch(batchApexInstance,batchSize,Prioity,isRetry); AsyncApexFramework.flush();
Where ,
- @Param 1 – Instance of Batch Apex
- @Param 2 – Batch Size / Scope Size
- @Param 3 – Priority to process if there are many job in Queue. If null then default would be 99
- @Param 4 – if Batch Apex has error, should it be retried again ? Use this option carefully , make sure your design does not negatively impact if same Batch runs multiple time
Same way , instead of using
Id jobId = System.enqueueJob(queueableClassInstance)
We would be using
Id jobId = AsyncApexFramework.submitQueueable(queueableClassInstance,priority,isRetry); AsyncApexFramework.flush();
Where,
- @Param 1 – Instance of Queueable class
- @Param 2 – Priority to process if there are many job in Queue. If null then default would be 99
@Param 3 – if Batch Apex has error, should it be retried again ? Use this option carefully , make sure your design does not negatively impact if same Batch runs multiple time
Below sample code, shows how we can add 300+ job in single transaction using this framework
//Code snippet to submit Batch and Queuable
for(Integer i = 0; i<300;i++){
BatchDemo b = new BatchDemo('#B');
AsyncApexFramework.submitBatch(b,1,99,true);
AsyncApexFramework.submitQueueable(new QueueableDemo(),99,true);
}
AsyncApexFramework.flush();
Feel free to add your feedback and findings on this.
This is one solution for handling the Batch Jobs in Salesforce.
Having worked in Mainframe Jobs for some time, where it is not uncommon to have more than 1000 batch jobs running for an organization at any time, I think that providing a similar environment to run batch jobs becomes responsibility for Salesforce especially when many customers want to migrate their applications to Salesforce because of its GUI, Sales Cloud, Service Cloud Features, Ability to quickly develop applications. 100 jobs in FlexQueue is a good start. However this is not a sufficient number.
Salesforce should develop facilities such as Auto Scaling, Load Balancing etc to take care of similar situations (as provided by Amazon Web Services etc).
GUI is a strength of Salesforce. However, it appears like, it needs to strengthen batch capabilities.
Hey Guru, While i appreciate a different perspective from your MainFrame work experience but you cannot really compare AWS & Salesforce. For most of the enterprise customers, former being a more private cloud, compared to public one like salesforce. There is a reason, we have governor limits like on a multi-tenant platform like Salesforce. If you really want to compare AWS capabilities , then Heroku ( runs on AWS, surprise) would be a good platform.
Hey Tim,
You are right. We can not compare Salesforce with Mainframe or even AWS. I just mentioned,if some of the useful features in different domains could be added to Salesforce to make it more worthwhile.
Yes Salesforce is a far better platform than AWS and comes with lots of features and makes it more easy to use. Since it is a public cloud with lots of customers residing on the cloud, features like Governor Limit need to be there. Probably for different licence types, more relaxed governor limits need to be there.
Thanks for your reply
Pretty freaking cool, Jitendra!
This is great Jitendra, came across your blog for the first time and this is one of the cleanest and nice way to approach this problem.
Nice solution Jitendra! Great work.
this is exactly what I’m looking for. thanks for sharing.
Thank you so much Jitendra, where can I get the object definition for the AsyncQueue object, especially for the picklists etc?
You’re my hero, Jitendra.
I noticed that (if the flex queue is full) records were getting created with the Status and Retry Count fields as null. Modifying the code, or otherwise defaulting those fields to “Queued” and “0” respectively seems to solve the problem.
Thank you again for this awesome program. If you see this message, please let me know if I’ve made a mistake in my assessment.