You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Problem:
Cromwell on Azure deployer is creating VM and Storage Account resources using a naming scheme of "prefix + random digits". While this helps to create unique entities, it becomes a maintenance hassle once the resources are created and being maintained over the lifetime of the Cromwell server. Try typing in your random storage account a few dozen times without carefully looking at it or copy/pasting it.
For example, the deployer takes a prefix and then adds a different random UUID suffix to that name. This means that the UUID used in the VM is different than the UUID used in the Storage Account. Which makes it basically impossible to ever memorize those resource names, so either you have to go to the Azure Portal to copy/paste the names or you have to create aliases you can use (i.e., environmental variables).
Solution:
It would be better if the deployer created resources using either just the user provided prefix. Or that prefix plus something sensible or a dictionary style generator. For example, if the prefix was set to 'genomics-pipeline' you might generate something like:
VM: genomics-pipeline-vm or genomics-pipeline-vm1 if that name wasn't available
SA: genomicspipeline or genomicspipelinesa or genomicspipelinesabrowncat
Similarly for pool resources, the name should be more comprehensible. Pool names should be 'prefix + full vm family name + perhaps managed identity name, etc. + some small subset of unique digits'. So rather than: e575819cc52-D1_v2-UNZATQPZ44VEVK6QWJIDVROOMMSC26ZV-5CGYD5BFMSNKG
You would ideally want something like: wdltest-standard_d1_v2-wdltest.azurecr.io-5CGYD5BFMSNKG
or wdltest-standard_d1_v2-ConfigA-5CGYD5BFMSNKG wdltest-standard_d1_v2-ConfigB-QWJIDVROOMMS
The text was updated successfully, but these errors were encountered:
Problem:
Cromwell on Azure deployer is creating VM and Storage Account resources using a naming scheme of "prefix + random digits". While this helps to create unique entities, it becomes a maintenance hassle once the resources are created and being maintained over the lifetime of the Cromwell server. Try typing in your random storage account a few dozen times without carefully looking at it or copy/pasting it.
For example, the deployer takes a prefix and then adds a different random UUID suffix to that name. This means that the UUID used in the VM is different than the UUID used in the Storage Account. Which makes it basically impossible to ever memorize those resource names, so either you have to go to the Azure Portal to copy/paste the names or you have to create aliases you can use (i.e., environmental variables).
Solution:
It would be better if the deployer created resources using either just the user provided prefix. Or that prefix plus something sensible or a dictionary style generator. For example, if the prefix was set to 'genomics-pipeline' you might generate something like:
VM:
genomics-pipeline-vm
orgenomics-pipeline-vm1
if that name wasn't availableSA:
genomicspipeline
orgenomicspipelinesa
orgenomicspipelinesabrowncat
Similarly for pool resources, the name should be more comprehensible. Pool names should be 'prefix + full vm family name + perhaps managed identity name, etc. + some small subset of unique digits'. So rather than:
e575819cc52-D1_v2-UNZATQPZ44VEVK6QWJIDVROOMMSC26ZV-5CGYD5BFMSNKG
You would ideally want something like:
wdltest-standard_d1_v2-wdltest.azurecr.io-5CGYD5BFMSNKG
or
wdltest-standard_d1_v2-ConfigA-5CGYD5BFMSNKG
wdltest-standard_d1_v2-ConfigB-QWJIDVROOMMS
The text was updated successfully, but these errors were encountered: