SPYT Update Guide

This document outlines how to update SPYT using the Kubernetes operator.

Important Information Before Updating

During a SPYT update:

• The YTsaurus cluster continues operating normally.
• YTsaurus server components aren’t restarted.
• Data access via the YTsaurus API remains available.
• The update installs the new version alongside the current one without replacing it (Multiple SPYT Versions).

Before updating:

1. Verify compatibility between SPYT, Spark, Java, Scala, and Python versions.
   Refer to the compatibility table.

2. Review available SPYT versions and their changes.
   See the SPYT releases.

Preparation for the Update

1. Check the Current SPYT Version

Retrieve the current SPYT version from the specification:

kubectl get spyt -n <namespace> -o yaml | grep image

Example output:

image: ghcr.io/ytsaurus/spyt:2.8.0

2. Check SPYT Status

Verify that SPYT is in the Finished state:

kubectl get spyt -n <namespace>

Example output:

NAME            STATUS
<spyt-name>     Finished

If SPYT isn’t in the Finished state, wait for ongoing operations to complete.

3. Save a Backup of the SPYT Specification

Save a backup of the current SPYT specification:

kubectl get spyt -n <namespace> -o yaml > spyt-backup.yaml

Updating SPYT

The operator doesn’t modify existing SPYT resources. Instead, it creates a new version as a separate resource. Both versions coexist on the cluster — their artifacts are stored side by side in Cypress. You can delete the old version later.

1. Prepare the Specification for the New Version

Use the current SPYT specification as a basis:

kubectl get spyt <spyt-name> -n <namespace> -o yaml

Create a spyt-<new-version>.yaml file with:

• A new resource name in the metadata.name field.
• A new image version in the image field.

Example specification for version 2.9.0:

apiVersion: cluster.ytsaurus.tech/v1
kind: Spyt
metadata:
  name: spyt-2-9-0
  namespace: <namespace>
spec:
  image: ghcr.io/ytsaurus/spyt:2.9.0
  ytsaurus:
    name: <ytsaurus-name>

2. Apply the New Version

Apply the new version specification:

kubectl apply -f spyt-<new-version>.yaml -n <namespace>

The operator creates a new SPYT resource and uploads its artifacts to Cypress. The existing resource remains unchanged.

3. Track the Update Progress

Monitor the SPYT status during the update:

kubectl get spyt -n <namespace> -w

Update States:

4. Verify the Update

1. Confirm that both resources are in the Finished state:

   kubectl get spyt -n <namespace>
   

   Example output:

   NAME             STATUS
   <spyt-name>      Finished
   <new-spyt-name>  Finished
   

2. Check the new image version:

   kubectl get spyt -n <namespace> -o yaml | grep image
   

   Example output:

   image: ghcr.io/ytsaurus/spyt:2.9.0
   

Multiple SPYT Versions on a Single Cluster

The update process adds a new SPYT version as a separate resource, allowing multiple versions to coexist on the same YTsaurus cluster. Different teams can use different versions simultaneously.

SPYT and Spark artifacts are stored in Cypress at the path //home/spark/. The operator doesn’t delete old artifacts, so versions don’t interfere with each other.

Removing an Old SPYT Version

When the old SPYT version is no longer needed:

1. Delete the SPYT resource:

   kubectl delete spyt <old-spyt-name> -n <namespace>
   

2. Remove old artifacts from Cypress (if necessary):

   • Open the //home/spark/spyt/releases and //home/spark/conf/releases directories.
   • Locate nodes with the unwanted version number.
   • Delete them from the context menu