Fix Solr 10 migration

[dpol1]

Fixes #1831

[rzo1]

Hi @dpol1

thanks for the PR !

Can you check the pipeline status? Guess it needs auto formatting. Thanks

[dpol1]

Oops, my bad 😅

[rzo1]

For me, it looks good but I am not deep into SOLR.

[rzo1]

Thanks for the PR, @dpol1

[mvolikas]

The default configsets in Solr 10.0.0 use <luceneMatchVersion>10.3</luceneMatchVersion>, so we might as well pin to this version instead (minor).

[mvolikas]

Some options (like --conf-name) are not documented in the Solr Control Script Reference, but are shown when running bin/solr create --help so I guess are ok to use.

[mvolikas]

This wraps a synchronous request, which will only block a different thread.
That is the reason why we had used lbHttp2SolrClient.requestAsync which will start a "real" async network request to Solr. (see #1488 and the discussion here)

We could keep both the old logic and the new CloudSolrClient by doing the following:

1. When building the CloudSolrClient, specify a HttpJettySolrClient (like you do for the ConcurrentUpdateJettySolrClient) to ensure that the created LBSolrClient will be an LBAsyncSolrClient.

HttpJettySolrClient jettyClient = new HttpJettySolrClient.Builder().build();

CloudSolrClient.Builder builder =
        new CloudSolrClient.Builder(
                Collections.singletonList(zkHost), Optional.empty())
                .withHttpClient(jettyClient);

2. Get the async client before making a request.

LBAsyncSolrClient lbAsyncSolrClient = (LBAsyncSolrClient) ((CloudHttp2SolrClient) cloudClient).getLbClient();

[mvolikas]

cloudClient.getClusterState() is deprecated. We can use cloudClient.getClusterStateProvider().getClusterState() instead.