RollingUpgrade Opensearch

[SadDE]

Hi everyone!

Could someone please help me? Maybe someone has already encountered a similar issue?

Currently, my cluster consists of 10 nodes:

• datanode1 version 2.19.0

• datanode2 version 2.19.0

• datanode3 version 2.19.0

• datanode4 version 2.19.0

• datanode5 version 2.19.0

• datanode6 version 2.19.0

• datanode7 version 2.19.0

• datanode8 version 2.19.0

• coordinatornode version 2.19.0

• masternode version 2.19.0

I want to perform a Rolling Upgrade of my cluster to version 3.1.0 and maintain index availability.

However, after I start upgrading the first node:

• datanode1 version 3.1.0 <----- This node

• datanode2 version 2.19.0

• datanode3 version 2.19.0

• datanode4 version 2.19.0

• datanode5 version 2.19.0

• datanode6 version 2.19.0

• datanode7 version 2.19.0

• datanode8 version 2.19.0

• coordinatornode version 2.19.0

• masternode version 2.19.0

The indices receive “unassigned shards” status.

Question: Are indices from version 2.19.0 and 3.1.0 compatible within the same cluster? If yes, how can I achieve this?

Thank you!

[merlinz01]

Have you re-enabled shard replication after restarting the updated node?

PUT "/_cluster/settings?pretty"
{
    "persistent": {
        "cluster.routing.allocation.enable": "all"
    }
}

After you do that it can take a while until the shard are redistributed and the cluster status is green.

If that doesn’t work I would guess that the unassigned shards status is to be expected until all nodes are on the same major version.

Rolling upgrades are supported between adjacent major versions, in this case 2 and 3. Make sure you have backups in case something gets messed up.