Closed clement911 closed 4 years ago
@clement911 Thank you for reaching out. This is being assigned to the content owner to provide more insights.
@clement911 Yes, we are planning to support changing the service tier from Hyperscale to other service tiers. I do not have a timeline to announce at this point, but we do recognize the importance of this, and the fact that some customers are hesitant to adopt Hyperscale otherwise.
The approach that has worked for many customers so far is to create a copy of their database, migrate it to Hyperscale, and run representative workloads on this copy to test functionality and performance, just like they would do for any other migration to a different database platform/technology.
Thanks @dimitri-furman but for us it would be very hard to test our production workload on a copy of the db.
On the other hand, if we could, we would easily do the following. 1) Migrate to Hyperscale 2) Monitor for a few hours/days 3) Decide whether to stay on Hyperscale or revert to Business critical
Our database is about 2TB and growing fast, so we are very interested in Hyperscale, but it's too risky for us to migrate without a revert path.
Is there any plan to allow for a bacpac to be created?
@mrblonde91 It is possible to create a bacpac today for smaller Hyperscale databases, though not in the same way as for other Azure SQL databases. To quote from the main Hyperscale documentation article:
Bacpac export/import from Azure portal, from PowerShell using New-AzSqlDatabaseExport or New-AzSqlDatabaseImport, from Azure CLI using az sql db export and az sql db import, and from REST API isn't supported. Bacpac import/export for smaller Hyperscale databases (up to 200 GB) is supported using SSMS and SqlPackage version 18.4 and later. For larger databases, bacpac export/import may take a long time, and may fail for various reasons.
I see this was closed. I hope you make this a priority. We are extremely keen to test Hyperscale but can't because of this issue....
@clement911 Yes, we understand that this is a blocker for Hyperscale adoption for some customers, and we do treat this as a priority. (The issue was closed because there were no updates for over a month, please feel free to reopen at any time.)
Are there still no updates on this? We are very keen to try Hyperscale and migrate if all is green but we can't due to this issue.
@dimitri-furman could you please reopen this issue? I don't think I can.
@clement911 We continue working to allow migration from Hyperscale to other service tiers. While we have made good progress in recent months, I unfortunately cannot provide an exact time when this will become available.
It also doesn't appear that I can reopen this issue, but please feel free to leave additional comments.
Still no eta 🙏? That would be a great Christmas pressie 😉
The best ETA I can provide is a public preview of that capability in the first half of 2021. But this is only with medium confidence at this point, so customers should not make plans with the expectation of this being available at any specific point in time.
I'll take what I can get! Hopefully we don't quite reach 4TB by then
In exactly the same boat as clement911. We've done a PoC on the Hyperscale tier and are mostly satisfied with the tests. But that 1 limitation of not being able to come back to Business critical or another tier, is making us nervous to attempt permanently moving to Hyperscale. We never know what we might find after moving and so this limitation is a big one for us.
@Darryl284 FYI, we ended up taking the plunge and migrating to Hyperscale anyway because we were getting very close to the 4TB limit. It turned out OK and we had no issues but it was definitely very scary to do, especially the warm up time just after the transition where the DB was quite slow for a couple of hours, but things settled after that.
We do face a major issue not being able to rebuild some indexes, but I'm not sure whether it is not related to HyperScale or not.
Your workload may be different but I just wanted to explain my experience.
Thanks @clement911 , we're doing a proof of concept for analytical loads. So far it seems good but there is always that fear of missing testing something. Having the capabity to be able to move back to other tiers would have made this decision so much easier. Thanks for sharing your experience though.
I know what you mean. It is always really scary to not be able to revert. That's interesting to me. Our workload is hybrid OLTP and OLAP and we use replicas heavily for all our analytical queries.
Will it ever be possible to migrate from Hyperscale to an other tier?
We can't confidently try out Hyperscale without being able to revert to our current tier (Business critical) if something goes wrong or if performance is not acceptable.
We understood this limitation would be removed fairly quickly but nothing has happened for months.
Document Details
⚠ Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.