Skip to content

Migrate from Polars Cloud and Polars On-Prem

Polars Cloud and Polars On-Prem are now one product, Polars Nanook. Polars Cloud became the AWS deployment, and Polars On-Prem the Helm chart and bare-metal deployments. Class and function names stay the same, and the old names below keep working for now. A future release removes them, so switch soon.

Client

Old New
Package polars-cloud polars-nanook
Polars extra polars[polars-cloud] polars[polars-nanook]
Import import polars_cloud as pc import polars_nanook as pn
CLI pc nanook
Environment variables POLARS_CLOUD_<NAME>, such as POLARS_CLOUD_CLIENT_ID NANOOK_<NAME>, such as NANOOK_CLIENT_ID
Config directory polars_cloud polars_nanook

Polars' own POLARS_CLOUD_ variables for cloud storage, such as POLARS_CLOUD_MAX_RETRIES, keep their names. The client moves the config directory when it starts. The polars_cloud import and the pc command only work while the polars-cloud package is installed.

Logins from cloud.pola.rs no longer work, so log in again with nanook login. Service accounts keep working. Restart any AWS cluster that was started before the move to nanook.pola.rs.

Helm chart

The container image moved from polarscloud/polars-on-premises to polarsinc/polars-nanook-compute on Docker Hub, and earlier versions were copied over. New versions are only published to polarsinc, and later chart versions use it by default.

Bare-metal

Old New
Package and command polars-on-premises, from a private package index polars-nanook-compute, from PyPI
Binary polars-on-premises-<VERSION>-linux-<ARCH> polars-nanook-compute-<VERSION>-linux-<ARCH>, from cdn.onprem.pola.rs
Environment variables PC_CUBLET__<KEY> NANOOK__<KEY>

Hosts

Old New
Portal cloud.pola.rs nanook.pola.rs
Documentation docs.cloud.pola.rs docs.nanook.pola.rs
API api.prd.cloud.pola.rs api.prd.nanook.pola.rs
Login auth.prd.cloud.pola.rs auth.prd.nanook.pola.rs
Clusters *.compute.cloud.pola.rs *.compute.nanook.pola.rs

If your network allowlists outbound hosts, add the new ones. The old hosts stay up for a while, so keep them allowed until every client and cluster is upgraded.