E2 custom shared-core instance pricing, mistake?

[sam_melrose]

Hello,

I've noticed what I think is a mistake in the E2 custom shared-core instance pricing.

I switched from an e2-micro to an e2-custom-small-4096 in a project with just that single VM (testing project).

My billing for SKU 9FE0-8F60-A9F0 went from 6 vCPU hours per day, to 48 vCPU hours per day.

I believe an e2-small should be 12 vCPU hours per day, so it is incorrectly charging me for 2 vCPUs (instead of 0.5).

Has anyone else spotted this? And is anyone from the product team able to comment?

Regards,

Samuel Melrose

[AndrewB]

• Hi @sam_melrose , have you contacted our Billing team yet? If you haven’t, feel free to reach out to them here https://cloud.google.com/support/billing/

[sam_melrose]

I’ve had a ticket open for about 48 hours, but they’ve marked the priority down and don’t seem to be getting back to me (logged from our gold support plan).

I was hoping this might be a more direct way to reach the product team while this product is in BETA

[Moazzam]

Hi Sam, Can you share the ticket with me - you can email me at moazzamchaudry@ and we will look at it.

[sam_melrose]

Thanks Moazzam,

Support replied to my ticket shortly after my last post here to say they were taking it seriously and had passed it on, but I will also drop you an email with the ticket details.

[sam_melrose]

I'm afraid I spoke too soon, support have replied and are trying to argue that an e2-custom-small-4096 is correctly being billed at 2 vCPU / hour.

I don't understand their logic, why would you choose it over an e2-custom-2-4096 if they are billed the same?

They are essentially saying with an E2 shared-core custom instance, you are paying the same as a more premium 2 vCPU instance to artificially reduce your own CPU allocation - what is the point?

This must be a misunderstanding, as otherwise it's a scam product designed to mislead.

[sam_melrose]

This has been escalated to the top level of our Gold support, to the escalation manager and it’s a P0 - Critical…

But they still refuse to believe an e2-custom-small-4096 shouldn’t cost the same as an e2-custom-2-4096

It’s honestly beyond belief the lack of common sense.

I’ve escalated to our account management team to try and get us in touch with the product team directly, will update on progress.

[dcotruta]

I've been looking at this type of machine: did you ever resolve this?

[sam_melrose]

Apologies for the delayed reply @dcotruta ,

We had heard at the beginning of last week that they had confirmed the problem and were investigating, so it's great to have an on the record confirmation from @Shamel , much appreciated! :).

Massive shout out and thank you to our Customer Engineer, Nick (not sure if I'm allowed to provide his surname): he spent a lot of time chasing this down internally and from what I can gather, it was pretty painful getting someone to acknowledge it, so thanks Nick!

Still no word from the product team about bill credits for those of us who have been massively over-billed for using this instance type over the last few months, but our account management team have said they'll be pushing for that once the fix has been implemented.

[Shamel]

After further investigation, it was determined that the e2 custom shared core instances were in fact overcharging for its use. The fix is in progress and will correct the pricing to charge for the fractional vCPUs, which are 0.25 (micro), 0.5 (small), or 1 (medium), instead of the full 2 vCPU. As mentioned in the documentation, these instances are subject to the same pricing rate as E2 custom machines (E2 custom vCPU and custom memory rates)

https://cloud.google.com/compute/vm-instance-pricing#e2_sharedcore_machine_types

We will update this thread with next steps once the updated pricing is available.

[TaradinoC]

Any update on this? I've just run into the same issue with one of my instances.

[Shamel]

This issue has been resolved as of December. Are you seeing this issue in which an E2-custom-(micro|small|medium) instance is still being billed at 2 vCPU?

[TaradinoC]

Yes, I migrated an instance from e2-small to e2-custom-small-3072 and was surprised to see it being billed as 2 vCPU instead of 0.5 vCPU.

Before the migration, I had two e2-small instances running, and their usage was covered by a 1 vCPU commitment. After the migration, with one e2-small and one e2-custom-small-3072 instance running, the commitment was still fully utilized and I was also being charged the on-demand price for another 1.5 vCPU.

I have a support case open; I'll link them to this thread.

[sam_melrose]

I'm still seeing this issue and still haven't been offered any bill credits for the ~5 months I've been billed incorrectly.

I have to say I'm really disappointed about this.

[Shamel]

We've verified that the fix for this issue has taken effect as of May 11th. Can you verify that the correct cost for the e2 custom shared core instance is reflected in your project billing for May 11th onward?

There will be an adjustment provided to compensate customers for the overcharge amount incurred from these instances. We will be in contact soon on when the adjustment. We understand that the complete fix for this issue has taken longer than expected and apologize for any inconvenience.

[Shamel]

A billing adjustment is being processed, and an email notification has recently been sent in regard to this issue. Please keep a look-out for the email and the adjustment that will be recognized on a future invoice. Thank you,

[jbogusz]

I might be missing something, but VM creation wizard seems to be in conflict with documentation. Is this an issue that will be fixed or am I missing something here?

[alexmoore]

e2-small has 2GB of memory and is presented as 2 vCPUs with a fractional capacity of 0.5 vCPU, see: https://cloud.google.com/compute/docs/general-purpose-machines#e2_machine_types