We only have a execute stored procedure in ADFv2. But most of the time we don't want to create stored procedure for all of the primary ETL tasks, such as counting the no. of records from a table, Updating data into tables, creating tables, etc. There are many such activities need T-SQL Execution. It would be great if we have Execution SQL option.

ADFv2 have the option to use variety of RDBMS source and sink systems such as MySQL, Oracle, etc., . ESQL would be the powerful task to have in Azure Data Factory V2 to be used in all of the RDBMS source / sink systems, as SQL is compatible for all of them.

Some of the other points to have ESQL task is highlighted,

1. Stored procedures are not the effective object to understand, debug complex sql code written within. With E-SQL, a large complex SP could be split at many places in the ETL flow and graphically represented to easy understanding. - Graphical representation is the core feature for any ETL tool.

2. Version management - For example. A create table script could be placed on a blob storage. which can be used as a source for E-SQL task. This helps to manage the table object in terms of adding additional columns in future could be done from blob / file storage

3. When SSIS is rebuilt on Azure Data Factory (which is the ultimate goal for Azure Data Factory V2). ESQL is used quite commonly in SSIS. By having this in ADF, it helps for quicker development

Additional feature with E-SQL could be great is to use source and sink systems. For example, getting data from Oracle source system and which need to perform a lookup with SQL Sink, and load them back to SQL Sink. If we have the option on ESQL to use both source and sink. It would be more powerful.

Thanks and Regards,
Vijay

Static Public IP's cannot be used with Virtual Network Gateways. This can potentially be very problematic if a Virtual Network Gateway ever needs to be re-created or re-provisioned.

Example: what if we have 30 separate tunnels to a Virtual Network Gateway and it needs to be re-created or re-provisioned? This would result in a new Public IP being provisioned (takes about 30-40 minutes - of downtime!) which would require 30 remote VPN Administrators to be engaged to rebuild their side of the tunnel. This could be easily resolved by allowing Static Public IP's to be associated with Virtual Network Gateways.

Furthermore, if this feature is implemented, it should be done in such a way that allows existing VNG's to set their Public IP to static WITHOUT requiring the IP to change. Example: If we have 20 VNG's in 4 separate regions when you implement the feature, it would not even be feasible for us to use the feature if we will lose our existing IP when setting it to static.

Prevent Creation of Legacy Machine Learning Workspace Commitment Plans and classic web services so Tags can be Enforced.

Our organization sets a Tag policy on resource groups so that newly created resources within that resource group will inherit tags. The main tag is 'Department Code', and is used for Chargeback for resource usage. Any Machine Learning 'classic' web services, that were created using the Machine Learning Studio portal will be created outside of the standard tagged resource group and thus, will not inherit standard tags used for Chargeback.

We also need a migration method for existing classic web services to be migrated to ARM web services, which is currently not supported by Microsoft. We will need to delete deployed instances from classic web services and deploy web services instead of classic web service from ML studio, and this has to be done one experiment at a time.

Recently we had a problem where we had ~1h of downtime which could have been avoided if VIP swap wasn’t failing.
The problem was that 1 single VM was in a busy state in Production slot. All VMs in the staging slot were up and Running.

Can you please allow a VIP swap if:
1. All VMs in Staging slot are in Running state
2. There’s the same or higher number of VMs in Staging slot compared to Production

I know that it’s by design that VIP swap is possible only if all VMs in both Staging and Production slot are in Running state. However, the point of doing the VIP Swap is to promote Staging slot to Production quickly, without downtime and to recover quickly from a bad state in Production slot to a healthy one in Staging slot.

Having a single busy VM in the Production slot, blocking the recovery and causing unnecessary service downtime does not make sense and looks like horrible design.

CSP is a great purchasing model for small and medium businesses, but has some serious problems.

1) Migrating between CSP and any other offer is a very laboursome and expensive project (not to mention that many resources cannot be migrated). This can cause Vendor lock-in and some other major risks for customers using CSP. Currently even PAYG to CSP is not easy.

2) CSP customers are unequal to other Azure users despite the fact that they pay the same or sometimes even more, because they do not have the same services available to them.

3) CSP resellers have to put in a lot of labour to onboard existing customers.

Please work on harmonizing the services between the different offer types so that a Customer in CSP is equal in terms of what services they can use. Furthermore, work on enabling (MS Support assisted or self-service) migration from CSP to PAYG/EA AND vice versa ON THE BILLING LEVEL without service interruptions (as with PAYG to EA or EA to PAYG).