I found in here that a tool is being worked on to do a migration from Elastic to OpenSearch.
This will however require that both are running in parallel.
In our scenario we have two geo-redundant sites with each a cluster of 3 Windows nodes.
One of the nodes in each site is the DataMiner server where an additional disk is installed for Cassandra and one for Elastic.
DataMiner is currently offloading to both Elastic clusters.
If data needs to be maintained we either need to install OpenSearch on the same servers in Windows or set up new servers. That last is not practical since it would require 6 new servers to be installed while the 'old' 4 are actually not old at all.
Is following possible, or is there another method that is being advised in this scenario?
- uninstall Elastic on backup site
- I would prefer to remove the search db on the DataMiner server.
Add one server to replace the DataMiner server node - Install Ubuntu server on all 3 backup site servers
- Install OpenSearch on backup site
- Do migration from main site (Elastic) to backup site (OpenSearch)
- point DataMiner to the backup site OpenSearch
- Set up additional server on main site
- uninstall Elastic on main site
- set up ubuntu server on 3 main site servers
- install OpenSearch on main site
- Have DataMiner offload to both sites.
During this switch there will be no geo-redundant backup, but there's no risk of data-loss.
Hi Jeroen,
The procedure is feasible as each of the two Opensearch clusters are treated independently. However this would require downtime of the DMS from step 5 onwards up until step 11. This to ensure that the geo redundant OpenSearch clusters do not deviate from each other in available data.
Running a DMS with a geo-redundant indexing cluster (1 Elastic & 1 OpenSearch) will not work, further requiring the system to be down until both geo-redundant OS clusters are up & running.
Is it the advised procedure for this scenario?
What would be the downtime in case we go with this procedure?