Automated: Migrating Ceph Data Pool From Replicated to Erasure-coded Type

Overview

To reduce the Objectstore storage requirements for the HA deployment model, you can migrate Ceph data pool from replicated type to erasure-coded type.

In versions prior to 2022.4, Ceph was using the replicated type of data pool, which takes 900 GiB of space across the cluster to store 50 GiB of objects. The erasure-coded pool will reduce the storage space required to store the same 50 GiB objects from 900 GiB to 450 GiB.

Note: The Ceph data pool migration is not required for non-HA deployments.

Prerequisites

Automation Suite deployed in HA mode is required for the Ceph data pool migration.

Migration Steps

To migrate Ceph data pool from the replicated to erasure-coded type, take the following steps:

1. View the help for the migration script by running the following command:

   sudo ./Tools/rook-ceph-data-pool-migration.sh --help

   Output:

   ***************************************************************************************
   
   Migrate Ceph data pool from replicated to erasure coded pool
   
   Arguments
     -l|--log                            Log file path to store actual ceph pool migration logs
     --force                             Provide consent to perform action which may cause downtime to complete migration
     -h|--help                           Show this help message and exit
   
   ***************************************************************************************

2. Run the migration script with tmux /screen /nohup if your SSH session gets disconnected frequently.

   sudo ./Tools/rook-ceph-data-pool-migration.sh

3. Update the ArgoCD configuration by running the following command:

   kubectl  -n argocd get application  fabric-installer -o json | jq 'if ([.spec.source.helm.parameters[].name] | index ("global.rook.dataPoolType")) == null then .spec.source.helm.parameters +=  [{"name": "global.rook.dataPoolType" , "value": "erasure-coded"}] else (.spec.source.helm.parameters[] | select(.name == "global.rook.dataPoolType").value) |= "erasure-coded" end'  | kubectl apply  -f -

Note:

The script will perform the pre-migration checks to verify if the Ceph cluster is healthy and supported for the migration, and will only then perform the actual migration.

If the migration script fails with the Insufficient Storage space available for LH node error, and it is not possible to increase the storage space, you can perform the manual migration using Rados Export/Import. For instructions, see Manual: Migrating Ceph data pool from replicated to erasure-coded type.