When do asynchronous jobs fail, suspend or retry?

 

Gonzalo Ruiz wrote an excellent article on When do asynchronous jobs fail, suspend or retry? When the CRM Asynchronous Processing Service gets an error, there are three possible actions it will take depending on the type of error:

• Fail: Job cannot be resumed.
• Retry: Job will be paused and retried after a period of time.
• Suspend: Job will be suspended until it is manually resumed.

The entire error handling mechanism is rather complex but I thought of writing down some general rules that will help understanding what the outcome action will be depending on the error that occurs inside the asynchronous job:

Scenario	Async job result action	Workflow result action	Error code
An SDK call fails: Infinite loop detected	Fail		80044182
An SDK call fails: Organization disabled	Retry		8004A104 / 8004A107
An SDK call fails: Server is busy	Retry		8004A001
An SDK call fails: Other	Fail	Suspend
SQL exception is thrown	Retry		80040216
Workflow system is paused	N/A	Suspend	80045017
Network error	Retry		80044306
Record associated with workflow cannot be found	N/A	Suspend	80045031
The HTTP response fails with code HttpStatusCode.Unauthorized	Suspend		80044306
The HTTP response fails	Retry		80044306
Plugin or workflow activity throws an InvalidPluginExecutionException	Fail		80040265
Anything else	Fail

Please note that the table above can be used as a general guide but might not cover all scenarios, some exceptions to these actions might apply and it might go outdated in the future.

Why do workflows have a different behavior than other asynchronous jobs when an SDK call fails?

Because the user might be able to fix the problem and resume the workflow. For example, a workflow step sends an email to an account. If the account has no email address, the workflow will suspend with error message "This message cannot be sent to all selected recipients. The e-mail address for one or more recipients is either blank or not a valid e-mail address".

The user can add the email address to the account and resume the workflow. The reason why other asynchronous jobs fail instead is because while workflows are sometimes manipulated by end users, other asynchronous jobs are more oriented towards the system administrator or customizer.

When the result action is Retry, for how long will the job pause before automatically retrying and how much time is there between retries?

The "PostponeUntil" attribute of the asynchronous operation corresponds to the next time it will be retried. The “PostponeUntil” attribute can be retrieved using the SDK. The amount of time to wait until the next retry is calculated considering some deployment settings and grows exponentially on the number of retries. The calculation uses a complex algorithm but these are some default outputs as a function of the RetryCount (the number of times the operation has been retried before):

RetryCount	Time to wait (seconds)
0	36
1	43
2	52
3	62
4	75
>= 5	Suspend

Note that by default, any asynchronous operation retrying 5 or more times will be suspended.

When a job has statusreason "Waiting" and statecode "Suspended", how do I know if it will retry or if it is suspended until it is manually resumed?

You can check the "PostponeUntil" attribute of the asynchronous operation to see the time and date in which it will be automatically resumed. If this value is equal to 9999-12-30 23:59:59 (maximum DateTime value) it means that it is waiting to be manually resumed.

How can I retrieve the “PostponeUntil” attribute of the asynchronous operations?

Because the “PostponeUntil” attribute is not available from the entity form or advanced find, you will need to use the SDK to retrieve this value. The following code sample prints the date and time at which each suspended asynchronous job will resume and the number of times it has been retried.

   1: static void Main(string[] args)

   2: {

   3: CrmAuthenticationToken token = new CrmAuthenticationToken();

   4:     token.AuthenticationType = 0;

   5:     token.OrganizationName = "AdventureWorksCycle";

   6:     CrmService service = new CrmService();

   7:     service.Url = "http://crmserver/mscrmservices/2007/crmservice.asmx";

   8:     service.CrmAuthenticationTokenValue = token;

   9:     service.Credentials = System.Net.CredentialCache.DefaultCredentials;

  10:  

  11:     QueryByAttribute query = new QueryByAttribute();

  12:     query.Attributes = new string[] { "statecode" };

  13:     query.ColumnSet = new ColumnSet(new string[] { "postponeuntil", "retrycount" });

  14:     query.EntityName = EntityName.asyncoperation.ToString();

  15:     query.Values = new object[] { (int)AsyncOperationState.Suspended };

  16:     BusinessEntityCollection bec = svc.RetrieveMultiple(query);

  17:     foreach (BusinessEntity be in bec.BusinessEntities)

  18:     {

  19:         asyncoperation op = (asyncoperation)be;

  20:         string result = String.Format("Operation id={0} PostponeUntil={1} RetryCount={2}",

  21:              op.asyncoperationid.Value,

  22:              op.postponeuntil.UniversalTime,

  23:              op.retrycount.Value);

  24:         Console.WriteLine(result);