General

This page provides a summary of recommended best practices for managing your own deployment of VIAME-Web in GCP in the NOAA Fisheries Cloud. Note that these admin options should be evaluated and only used if appropriate to your specific use case.

Addon Management

To download and scan for VIAME addons, such as canned pipelines and algorithms, see the Addon Management docs.

Staying up to date

Keeping your deployment up to date is crucial for getting the latest VIAME-Web features and bug fixes, as well as for debugging purposes. It is strongly recommended that you update your images at least once per week.

The startup scripts that are created on your VMs during deployment provisioning include docker-compose pull calls that will update the containers installed on your VM(s) to the latest images. Thus, these containers will be updated every time you run the startup script. Depending on your VM management strategy this may be sufficient, particularly if you turn off your VM(s) at least once per week.

If your VM (default or web) will be on at all times, you should consider the Production deployment options.

Note that keeping images up to date will not update canned VIAME addons. See Addon Management for details on updating VIAME addons.

Data backup

Data in GCS Storage buckets are quite durable. See the GCS Storage docs for more details about data redundancy, versioning, etc.

You can create a GCP Snapshot Schedule to create snapshot backups of your disk(s).

Click here to download a Terraform code template and usage instructions for data backup and VM management. Copy this code into your Terraform config file (e.g., main.tf) and update project-specific values as needed. In the Cloud Shell Terminal, run terraform init, and then terraform apply to create the resources.

VM management

Your VM schedule, i.e., when your VMs are on (and incurring more costs) and when they are off, will be up to your group. You can use the Google Cloud Pricing Calculator to estimate costs.

If you have a split services deployment, an example schedule might be leaving the web VM during the work week, with an Instance Schedule set to turn it off on Friday evening, and turning the worker on and off as needed to run jobs. This would minimize users needing to turn on the web VM (or turning it off while another user is annotating), which also only having the worker VM on when necessary. Recent feature additions allow all users to see the status of the job queue so they know if someone else is currently running a job.

Click here to download a Terraform code template and usage instructions for data backup and VM management.

Server Branding Config

For information on how to configure the brand and messaging that appears in various places in the Web UI, see the Server Branding Config docs.

Storing imagery

Within GCP, storage on Compute Engine disks is much more expensive than storage in GCS buckets. Thus, it is recommended to store your imagery in a GCS bucket and mirror that bucket in your VIAME-Web deployment as a read-only assetstore. See Cloud Storage Integration.