Forge queue naming and pipeline failover

Forge creates AWS Batch queue and CE names using the format TowerForge-bunchanumbers-head, etc. We have pipelines using fallback strategies which depend on knowing the AWS Batch queue names (e.g. if spot gets pre-empted, move to the ondemand queue). But we can’t intuit what Tower will name the CEs, so each time we create new ones we have to update our Nextflow pipeline code. How should we be doing this?

The name follows a strict schema based on the provider, for AWS Batch it is ${TOWER_FORGE_PREFIX}-${computeEnvId}, with an additional suffix of -head or -work when using spot machines. By default, the prefix is TowerForge, so the resulting compute environments are:

TowerForge-${computeEnvId}-head
TowerForge-${computeEnvId}-work

We have a related open feature request for this here, where you could select a compute environment via the Seqera Platform name instead of the cloud provider ID. Does this sound like a suitable solution for you?

Thanks, Adam. It’s asking for a verified seqera.io email address to proceed. I work for Vertex Pharma, we have a site install of Tower, I’m not sure how to go about this.

My apologies! It’s been flagged as an internal request. I will get it moved and update here.

The feature request is basically you could use a config like this:

process {
    queue = 'my-compute-env-name'
}

and Seqera Platform would replace it with this under the hood:

process {
    queue = 'TowerForge-${UUID}-work'
}

Feature request has been moved to public facing, you can find it here:

That would be ideal. NAturally I’m wondering what the ETA is on that… Thanks!

Hi Jim - please upvote the feature request. Our Product team review all user-submitted feature requests weekly and customer upvotes (among other criteria) help inform our product roadmaps and priority. Additionally, if you upvote you will automatically get an email notification when the request changes state. Thanks!

I will be voting early and often! Thanks again.