Kai Walter
Kai Walter

Reputation: 4041

Why does Azure App Service Environment do the long deployment again when running the same ARM template?

I used the quick start template 201-web-app-ase-ilb-create to start creating my ASE (in West Europe). I let it run overnight. At some point in time it timed out. The next morning - looking at the portal - the deployment seemed to have succeeded. Before continuing adding the certificate for the ILB with the quick start template 201-web-app-ase-ilb-configure-default-ssl I reran the first template (just to be sure). Though no obvious change to the configuration is induced by the first template, a long deployment started again - probably for some more hours.

Upvotes: 3

Views: 6014

Answers (3)

RuSs
RuSs

Reputation: 1812

Regarding the "access token expiry" when deploying from Visual Studio, the deployment can still complete in Azure even if Visual Studio tells you it has timed out. Basically, you have already sent the ARM template to Azure so the time out just means the polling info about the deployment in VS stops working.

Upvotes: 0

bytejunkie
bytejunkie

Reputation: 1033

I've also seen some really long running ASE deployments. usually with a fail at the end. I found two issues were common to these long deployments.

  1. Most common was the fact i had forgotten the pre-requisite to have ports 454 and 455 open to the ASE subnet from all to all. this is what allows azure to do its magic and uses certs for security. if you're seeing over 2 hours for the deployment, make sure you've got this rule in place in your NSG.

  2. I'd got a dependson block in the wrong place, which caused a long long wait.

Upvotes: 2

Paco de la Cruz
Paco de la Cruz

Reputation: 2154

I have seen this behaviour every time we deploy ASEs using ARM Templates. Even though the Front-End and Worker Pools, App Service Plans and App Services don't change, they still take a couple of hours. What we usually do is that we use Incremental deployments and have the ASE in a different template. This is not ideal, but we haven't found any other way to avoid long-running deployments.

HTHs

Upvotes: 1

Related Questions