deploying haystack app via cloudrun w/ pgvectorstore

[M-JULIANI]

tl;dr:

any guidance on how to deploy a haystack rag pipeline (pgvector) to gcp cloudrun? running a flask app locally where i am able to successfully run indexing/inference pipelines, but struggling to do so via my cloudrun service.

hypothesis

while this may not be a haystack-related problem i am facing, it is worth asking whether this has been tried before and what the best practice is.

• i have a cloudrun container
• i have a cloud sql instance
• they are both connected via the environmental settings of my cloudrun container
• cloud proxy sql should be taking care of their connectivity per the docs.
• all permissions are set to public, and no authentication is required to access my cloudrun service
• (reference)

the error i am seeing in my cloudrun service logs are as follows:

insertId: "6633afc40002bfde4c944ee8"
labels: {1}
logName: "projects/spaice/logs/run.googleapis.com%2Fstderr"
receiveTimestamp: "2024-05-02T15:22:44.519715592Z"
resource: {2}
textPayload: "psycopg.OperationalError: connection failed: server closed the connection unexpectedly"
timestamp: "2024-05-02T15:22:44.180190Z"

there are a number of others, but this seems the like culprit.

wondering how this relates to the PG_CONN_STRING which haystack pgvector requires, versus what I am supplying (I am supplying both PG_CONN_STRING and the connection information that you can see in the tutorial I provided earlier).

any input on the topic is much appreciated.

[M-JULIANI]

Closing this discussion, once I got my connection string configured correctly between cloudrun and cloudsql, was able to run the whole thing just fine!