We received an Azure notice saying our Storage and Blob accounts must move from GPv1 to GPv2 before the October 13 automatic migration deadline. The portal presents this as a simple in-place upgrade with no downtime or data loss, but GPv2 changes the billing model: storage may be cheaper while transaction charges can be significantly higher.
For those who have already migrated, did you review transaction activity first or simply upgrade the accounts? If you analyzed the data, which Azure Monitor metrics or transaction levels influenced your decision? Did anyone experience unexpected cost increases, particularly with queue-heavy workloads, Azure Site Recovery, access tiers, or older ZRS configurations?
I'm trying to decide whether a detailed transaction audit is worthwhile or whether this is generally a routine, cost-neutral migration.
4 Answers
Our experience was mostly cost-neutral or slightly cheaper, but the storage accounts used for Azure Site Recovery became more expensive. I’d specifically separate those from ordinary blob storage when estimating the impact instead of assuming every account will behave the same way.
For most accounts, the migration really was just selecting the upgrade option and saving. We checked current transaction activity first and didn’t see a meaningful cost change after upgrading. I’d still pay extra attention to workloads that generate lots of small operations rather than large data transfers.
We built a script that reviewed 30 days of transaction metrics by transaction type and estimated GPv1 versus GPv2 costs for every account. Most accounts were close to their existing spend, so we upgraded them without redesigning anything. That kind of quick cost comparison is probably worth doing before pressing the button.
Some of our remaining GPv1 accounts were old and barely used, with tiny recurring charges left behind by previous teams. We mostly upgraded the ones that still mattered and deleted obsolete accounts. GPv2 also gives you lifecycle policies and access-tier options, which can be useful beyond the migration itself.
Queue-oriented workloads can be the exception—transaction charges may be dramatically higher there, so I wouldn’t skip the review if an account supports queues.