Cloud SQL Proxy

The Cloud SQL Proxy allows a user with the appropriate permissions to connect to a Cloud SQL database without having to deal with IP whitelisting or SSL certificates manually. It works by opening unix/tcp sockets on the local machine and proxying connections to the associated Cloud SQL instances when the sockets are used.

To build from source, ensure you have go installed and have set GOPATH. Then, simply do a go get:

go get github.com/GoogleCloudPlatform/cloudsql-proxy/cmd/cloud_sql_proxy

The cloud_sql_proxy will be placed in $GOPATH/bin after go get completes.

cloud_sql_proxy takes a few arguments to configure:

• -fuse: requires access to /dev/fuse as well as the fusermount binary. An optional -fuse_tmp flag can specify where to place temporary files. The directory indicated by -dir is mounted.
• -instances="project1:region:instance1,project3:region:instance1": A comma-separated list of instances to open inside -dir. Also supports exposing a tcp port instead of using Unix Domain Sockets; see examples below.
• -instances_metadata=metadata_key: Usable on GCE only. The given GCE metadata key will be polled for a list of instances to open in -dir. The format for the value is the same as the 'instances' flag. A hanging-poll strategy is used, meaning that changes to the metadata value will be reflected in the -dir even while the proxy is running. When an instance is removed from the list the corresponding socket will be removed from -dir as well (unless it was also specified in -instances), but any existing connections to this instance will NOT be terminated.

Note: -instances and -instances_metadata may be used at the same time but are not compatible with the -fuse flag.

By default, the proxy will authenticate under the default service account of the Compute Engine VM it is running on. Therefore, the VM must have at least the sqlservice.admin API scope ("https://www.googleapis.com/auth/sqlservice.admin") and the associated project must have the SQL Admin API enabled. The default service account must also have at least WRITER/EDITOR priviledges to any projects of target SQL instances.

Specifying the -credential_file flag allows use of the proxy outside of Google's cloud. Simply create a new service account, download the associated JSON file, and set -credential_file to the path of the JSON file. You may also set the GOOGLE_APPLICATION_CREDENTIALS environment variable instead of passing this flag.

Example invocations:

./cloud_sql_proxy -dir=/cloudsql -instances=my-project:us-central1:sql-inst &
mysql -u root -S /cloudsql/my-project:us-central1:sql-inst

# For -fuse you do not need to specify instance names ahead of time:
./cloud_sql_proxy -dir=/cloudsql -fuse &
mysql -u root -S /cloudsql/my-project:us-central1:sql-inst

# For programs which do not support using Unix Domain Sockets, specify tcp:
./cloud_sql_proxy -dir=/cloudsql -instances=my-project:us-central1:sql-inst=tcp:3306 &
mysql -u root -h 127.0.0.1

# Caution: This should be executed in a closed network. If executing on a system without a correctly configured firewall this could potentially allow anything on the internet to access the database.
# For accessing from another host in a network, specify host:
./cloud_sql_proxy -dir=/cloudsql -instances=my-project:us-central1:sql-inst=tcp:0.0.0.0:3306
# From another host:
mysql -u root -h [proxy-machine-ip]

To use inside a Go program:

If your program is written in Go you can use the Cloud SQL Proxy as a library, avoiding the need to start the Proxy as a companion process. If you're using the the go-sql-driver you can use helper functions found in the proxy/dialers/mysql package. See example usage.

I'm open to adding more drivers, feel free to file an issue.

To use from Kubernetes:

Kubernetes does not support the metadata server that is used by default for credentials, so we have to manually pass the credentials to the proxy as a Kubernetes Secret. At a high level, we have to create a Secret, add it as a Volume in a Pod and mount that Volume into the proxy container. Here are some detailed steps:

• Create a Service Account and download the JSON credential file, following these steps.
• Create a local Kubernetes Secret named sqlcreds from this file by base64 encoding the Service Account file, and creating a Secret file with that content:

apiVersion: v1
kind: Secret
metadata:
  name: sqlcreds
type: Opaque
data:
  file.json: "BASE64 encoded Service Account credential file."

• Create this Secret using kubectl create.

$ kubectl create -f secret.json

• Add the sqlcreds Secret in your Pod by creating a volume like this:

  - name: secret-volume
    # This GCE PD must already exist.
    secret:
      secretName: sqlcreds

• You'll also need to create a hostPath volume allowing the SQL proxy to read SSL certificates:

- name: ssl-certs
  hostPath:
    path: /etc/ssl/certs

• Create an emptydir volume named cloudsql for the SQL proxy to place it's socket:

  - name: cloudsql
    emptyDir:

• Add the SQL proxy container to your pod, and mount the sqlcreds and 'ssl-certs' volumes, making sure to pass the correct instance and project.

  - image: b.gcr.io/cloudsql-docker/gce-proxy
    volumeMounts:
    - name: cloudsql
      mountPath: /cloudsql
    - name: secret-volume
      mountPath: /secret/
    - name: ssl-certs
      mountPath: /etc/ssl/certs
    command: ["/cloud_sql_proxy", "-dir=/cloudsql", "-credential_file=/secret/file.json", "-instances=$MYPROJECT:MYINSTANCE"]

Note that we pass the path to the secret file in the command line arguments to the proxy. We also pass the project and Cloud SQL instance name we want to connect to using the "--instances" flag.

• To use the proxy from your application container, mount the shared cloudsql volume:

volumeMounts:
    - name: cloudsql
      mountPath: /cloudsql