Flush all redis instances after workflow #225
Labels
Comments
jcscottiii
added a commit
that referenced
this issue
May 9, 2024
This change modifies terraform to deploy redis. Given redis is a regional service and we are multi-region, we leverage the existing region-to-subnet map to know which regions to create redis in. In the meantime, we can have a quick TTL for redis. An issue to follow up on this in #225 has been filed. We can probably increase the TTL for prod soon. But while data is still changing before launch, we can keep it quick. Open to suggestions for the initial TTL. About the terraform changes itself: - Because the project spans multiple projects, we have a shared VPC. This is because Memorystore is different from other services like spanner. Spanner instances do not live in our IP space. But Memorystore instances do. As a result, we have to set up "Private Services Access" by giving it a slice of IPs for the service to deploy the service into. More here in the blue banner at the top of [1] and [2]. - The TTL for prod and staging are set to 5minutes. We can increase this in the future as we get closer to launch. Other changes: - Tell terraform to enable needed APIs. [1] https://cloud.google.com/memorystore/docs/redis/networking [2] https://cloud.google.com/memorystore/docs/redis/networking#private_services_access Change-Id: I47e8d7f5c8eee7f0e0ab90a2f8fe6c88bc79933e
dlaliberte
pushed a commit
that referenced
this issue
May 9, 2024
This change modifies terraform to deploy redis. Given redis is a regional service and we are multi-region, we leverage the existing region-to-subnet map to know which regions to create redis in. In the meantime, we can have a quick TTL for redis. An issue to follow up on this in #225 has been filed. We can probably increase the TTL for prod soon. But while data is still changing before launch, we can keep it quick. Open to suggestions for the initial TTL. About the terraform changes itself: - Because the project spans multiple projects, we have a shared VPC. This is because Memorystore is different from other services like spanner. Spanner instances do not live in our IP space. But Memorystore instances do. As a result, we have to set up "Private Services Access" by giving it a slice of IPs for the service to deploy the service into. More here in the blue banner at the top of [1] and [2]. - The TTL for prod and staging are set to 5minutes. We can increase this in the future as we get closer to launch. Other changes: - Tell terraform to enable needed APIs. [1] https://cloud.google.com/memorystore/docs/redis/networking [2] https://cloud.google.com/memorystore/docs/redis/networking#private_services_access Change-Id: I47e8d7f5c8eee7f0e0ab90a2f8fe6c88bc79933e
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
We should have a process that will proactively flush the redis instances when a workflow finishes.
The text was updated successfully, but these errors were encountered: