Mastodawn

David Chisnall (*Now with 50% more sarcasm!*)

Wow, moving an IP address between two #Azure subscriptions takes over an hour (and counting) and it has to be disconnected from a VM while doing so. How on Earth do you design infrastructure where an operation that should be updating one table in a database takes so long? Even manual telephone exchanges could do the equivalent operation faster.

David Chisnall (*Now with 50% more sarcasm!*)Feb 22

RE: https://infosec.exchange/@david_chisnall/116115244205622905

Wow, and #Azure support is terrible. It sends you to Copilot. Copilot says: 'If the resource group remains in the "MovingResources" state and you cannot identify any issues, it may be necessary to contact Azure Support. They can investigate backend issues and provide assistance in resolving the problem.'

Contacting support requires selecting from a drop-down list of common problems which does not include this problem. And there is no path I have yet found via their UI that actually seems to result in raising a support request.

Imagine trying to use this for important things!

David Chisnall (*Now with 50% more sarcasm!*)Feb 22

Oh, it gets better. Apparently I can't raise a support request even though their infrastructure is broken unless I add a support plan. And I can't add a support plan to this class of subscription.

What a clown show.

David Chisnall (*Now with 50% more sarcasm!*)Feb 22

Ugh. This sucks.

Azure decided to move my free subscription to a different subscription type. Fine, just log in and move it over to the other one.

Unfortunately, the move broke in the middle for some unknown reason. And there seems to be no possible way of contacting support.

Okay, so plan B: move it over to another machine and update DNS to point to it. Except #Gandi ( @gandi ?) now needs to send me an email to log into the DNS management console. And I can't do that because the DNS entry I want to update is the one to the mail server. So... any ideas?

They do have a recovery thing that is very complicated.

✧✦Catherine✦✧Feb 22

@david_chisnall please be aware that Gandi got bought out by private equity which did the thing everybody expects it to do (jacked up prices by at least 2x compared to equivalent registrars, and made services worse)

David Chisnall (*Now with 50% more sarcasm!*)Feb 22

Yup, moving off them is on the to-do list, but not the urgent thing for today. Getting working email again is the urgent thing.

Edwin Török Feb 22

@david_chisnall @gandi I just logged to my Gandi account and it didn't require receiving an email. I do have 2FA enabled with a Yubikey though.

They also have a "recovery by QR code", and "recovery by Security key" method. The QR code won't help if you never saved it, but if you ever added a Security key, then it might be possible to use that to recover.

David Chisnall (*Now with 50% more sarcasm!*)Feb 22

@edwintorok @gandi

That will probably help next time. Unfortunately, last time I logged in it didn't give me the option to set that up.

Steven Reed Feb 22

@david_chisnall
There's a reason @azureshit exists!

Dustin Mitchell Feb 22

@david_chisnall it's been a while but my impression of azure is it's meant for forklifting servers into the cloud and not dynamic usage. I suspect a lot of operations are shipped as an MVP where the API call creates a ticket and some operator does the work.

@djmitche @david_chisnall

IP-over-ticket !

lord pthenq1 Feb 22

@david_chisnall
Why are you using Azure?

Use gcp, AWS, linden (akamai), or digital Ocean.

David Chisnall (*Now with 50% more sarcasm!*)Feb 22

@pthenq1 Because they gave me a free subscription and I haven't got around to moving things off there yet.

But thanks for your completely unhelpful advice.

lord pthenq1 Feb 22

@david_chisnall
Oh. You are evaluating it.

I had bad experiences with Azure and Oracle.

Sorry I cannot help more

@david_chisnall (poorly designed) caching?

David Chisnall (*Now with 50% more sarcasm!*)Feb 23

I think the problem is that their migration is really recreating and moving the resources and has improper error handling. One of the properties exists only if explicitly set, so querying it returns an error. But the migration doesn’t stop, it just waits and retries (presumably because everything in Azure is so fragile that intermittent failures are normal). So now it’s stuck in a wedged state permanently.

@david_chisnall they need to get better vibes