Remove unassigned Graylog2 index/shards with Bash in a loop

Question

There were some disk issues on a Graylog2 server I use for debug logs. There are unassigned shards now:

curl -XGET http://host:9200/_cat/shards

graylog_292 1 p STARTED    751733  648.4mb 127.0.1.1 Doctor Leery 
graylog_292 1 r UNASSIGNED                                        
graylog_292 2 p STARTED    756663  653.2mb 127.0.1.1 Doctor Leery 
graylog_292 2 r UNASSIGNED                                        
graylog_290 0 p STARTED    299059  257.2mb 127.0.1.1 Doctor Leery 
graylog_290 0 r UNASSIGNED                                        
graylog_290 3 p STARTED    298759  257.1mb 127.0.1.1 Doctor Leery 
graylog_290 3 r UNASSIGNED                                        
graylog_290 1 p STARTED    298314  257.3mb 127.0.1.1 Doctor Leery 
graylog_290 1 r UNASSIGNED                                        
graylog_290 2 p STARTED    297722  257.1mb 127.0.1.1 Doctor Leery 
graylog_290 2 r UNASSIGNED 
....

It's over 400 shards. I can delete them without data loss, because it's a single node setup. In order to do this I need to loop over the index (graylog_xxx) and over the shard (1,2,...).

How do I loop over this (2 variables) with Bash? There are 2 variables for the deletion API call, which I need to replace (afaik):

curl -XPOST 'host:9200/_cluster/reroute' -d '{
        "commands" : [ {
              "allocate" : {
                  "index" : "$index", 
                  "shard" : $shard, 
                  "node" : "Doctor Leery", 
                  "allow_primary" : true
              }
            }
        ]
    }'

What also bothers me about this is, that the unassigned shards have no node. But in the API call I need to specify one.

Answer 1

Form the _cat/shards output you shared, it simply looks like those are unassigned replicas, which you can simply remove by updating the cluster settings and setting the replica count to 0, like this:

curl -XPUT 'localhost:9200/_settings' -d '{
    "index" : {
        "number_of_replicas" : 0
    }
}'

After running the above curl, your cluster will be green again.