Since catalyst servers are bound to a persistent volumes, a deployment schema where we delete the machine and create another (new) one is not a viable choice without downtime.
Instead, we will use the catalyst-owner/init.sh script to update the current version in every machine. To know where to update we will create two SNS topics that will be redirected to SQS queues (one for each deployment):
arn:aws:sns:us-east-1:619079673649:decentraland-catalyst-stable-version for
stable and tagged releases (official versions)
arn:aws:sns:us-east-1:619079673649:decentraland-catalyst-latest-version for
master branch versions (dev)
Every deployment will have a 1 minute crontab to pull messages of its own SQS
Create an SNS topic to publish new versions of the catalysts
In each catalyst deployment (EC2)
create also an SQS
subscribe that SQS to the SNS topic
crontab to consume the SQS in each deployment
sequenceDiagram
ci->>sns: new version published (using aws cli)
sns->>sqs: aws.sns.TopicSubscription
ec2->>ec2: cron (aws sqs receive-message)
sqs-->>ec2: message
ec2->>ec2: ./init.sh
ec2-->>sqs: delete-message (if succeed)
Properly receive SQS messages with aws cli
https://advancedweb.hu/how-to-use-the-aws-sqs-cli-to-receive-messages/
Create SQS with pulumi https://www.pulumi.com/docs/reference/pkg/aws/sqs/queue/
Subscribe SQS to SNS with pulumi: https://www.pulumi.com/docs/reference/pkg/aws/sns/topicsubscription/