How can we improve the Azure Resource Manager?

Application Gateway management operations are agonizingly slow

[Originally posted to Azure Portal forum - however should have been submitted here]

Most creates and updates for application gateway configuration (back-end pools, rules, listeners, certificates, etc.) take 5-10 minutes to apply. This is not a portal issue as it takes just as long via the API / CLI. (This is with even a single instance App Gateway)

Applying these same settings on Windows Server ARR directly takes only a minute or less, so I'm not sure where the performance issue lies, but there is tremendous room for improvement.

As it is currently, it provides waaaay too much time while waiting to consider the porting costs of a migration to the alternate cloud platform.

643 votes
Vote
Sign in
Check!
(thinking…)
Reset
or sign in with
  • facebook
  • google
    Password icon
    Signed in as (Sign out)
    You have left! (?) (thinking…)
    Kevin Piette shared this idea  ·   ·  Flag idea as inappropriate…  ·  Admin →

    59 comments

    Sign in
    Check!
    (thinking…)
    Reset
    or sign in with
    • facebook
    • google
      Password icon
      Signed in as (Sign out)
      Submitting...
      • guitarrapc_tech commented  ·   ·  Flag as inappropriate

        It takes15min+ to update listener's ssl certificate.
        The day passed now it's Autumn and will Winter comes?

      • Tristan Beasley commented  ·   ·  Flag as inappropriate

        Still really slow in West Europe, taken an hour so far to apply configuration changes to a Virtual Network Gateway, in the meantime it's down and unusable for 50 staff...

      • MikeN commented  ·   ·  Flag as inappropriate

        One suggestion I have for people complaining about this issue is if you have many different configuration changes to do, use PowerShell rather than the GUI. It seems that it's the saving of changes that takes a long time. You can make many different configuration changes, you only get the delay when you run set-azurermapplicationgateway to apply the configuration changes.

      • Johannes Edstoft Hansen commented  ·   ·  Flag as inappropriate

        For comparison our backend deployment consists of; AppInsights, ADB PostgreSQL, and an AKS cluster (with 5 nodes), which takes about 15 minutes to provision. AKS being the slowest resource to provision as it needs to provision the nodes vnets oms resources etc.

        However our frontend deployment consists of just an Application Gateway (Medium, 2 instances, WAF, minimum configuration, ~40 minutes to provision) and an API management (Basic, 1 instance, ~45-55 minutes to provision mostly spent "activating"). I feel both of these resources shouldn't take more than 10 minutes to provision each.

      • Johannes Edstoft Hansen commented  ·   ·  Flag as inappropriate

        THis problem also exists for API Management instances. Painfully slow to provision and change scaling on add SSL to etc. Please fix both of these! Testing ARM templates that contain either an APIM or AG currently takes > 90 minutes just in the provisioning phase (>45 minutes to bring up & >45 minutes to delete resources), this is wasteful and frustrating as what you typically want to test is the integration between resources.

      • MikeN commented  ·   ·  Flag as inappropriate

        Is there any update on this? It was working better for a while but now it is back to being ridiculously slow in US East.

      • Luke commented  ·   ·  Flag as inappropriate

        As others have mentioned change and creation times for Application Gateway (West Europe) are ridiculous, taking an hour or more just to front a single scale set. Not even with SSL.

      • Jason commented  ·   ·  Flag as inappropriate

        As an update, just created a basic App Gateway in West Europe, with a single front-end IP, a single listener (https with a single ssl certificate attached) linked to a single backend linux VM. The App Gateway took 52 minutes 30 seconds to build....this really needs work please, it's not acceptable. It almost forces you to keep fail-over gateways up at all times because the deploy time kills any SLAs you consider sticking to. Not good for the company wallet.

      • Jason commented  ·   ·  Flag as inappropriate

        This is still way too slow to be a useful product. The grating part is knowing that we are being charged for resources that take so long to come up, you know you've just paid money for that privilege of waiting, and when it does come up, or doesn't because of an error, you have to got through the same waiting game again! Please make this far more slicker!!

      • Amit M commented  ·   ·  Flag as inappropriate

        The times are simply ridiculous. Cannot wait for an hour to get the gateway created!!!

      • MikeN commented  ·   ·  Flag as inappropriate

        This seems to be much better now. Configuration changes are now saving quickly.
        Thanks for addressing this!

      • Anonymous commented  ·   ·  Flag as inappropriate

        Hi! Our team is going to use Application Gateway on production environment.
        When will it be fixed? Just simple deleting of listener takes an hour or more...

      ← Previous 1 3

      Feedback and Knowledge Base