NoNodes error if too many Catmandu::Store::ElasticSearch are created

[EreMaijala]

If you have a batch process that creates a new ElasticSearch store frequently, eventually searching using the newly-created store will fail with error "[NoNodes] ** No nodes are available". This is obviously sub-optimal, but it happens easily when you need to do an operation on a set of objects that use the ElasticSearch store internally.

Here is a very simplified sample script to reproduce the problem. On my Ubuntu 16.04.3 LTS box it fails after 1020 iterations.

#!/usr/bin/perl

use Catmandu::Store::ElasticSearch;

for ($i = 0; $i < 10000; $i++) {
    my $store = Catmandu::Store::ElasticSearch->new(
        nodes => ['localhost:9200'],
        index_name => 'test'
    );
    my $results = $store->bag->search(
        query => {
            match => {
                'title' => 'test' . $i
            }
        } ,
        limit => 10
    );
    print "$i\n";
}

[nics]

I'm going to close this ticket becausei don't think it's an issue with the Catmandu store. Probably related to something like connection timeouts. The store just passes settings like these to the underlying driver.