Cluster reroute api should have a way to assign all unassigned shards

[nik9000]

Sometimes, during a disaster, you end up with shards stuck unassigned and you just want them reassigned regardless of data loss. It'd be great if the cluster reroute api supported an assign_all action that just assigns all the shards to free nodes. It'd need to also support allow_primary. Without this you need some nasty shell script like this:

function reroute() {
    curl -XPOST 'localhost:9200/_cluster/reroute?pretty' -d '{
        "commands" : [ {
                "allocate" : {
                    "index" : "'$1'",
                    "shard" : '$2'
                    "allow_primary" : true,
                    "node" : "<node>"
                }
            }
        ]
    }' > /dev/null
    sleep 1
}
curl -s localhost:9200/_cluster/state?pretty | awk '
    BEGIN {more=1}
    {if (/"UNASSIGNED"/) unassigned=1}
    {if (/"routing_nodes"/) more=0}
    {if (unassigned && /"shard"/) shard=$3}
    {if (more && unassigned && /"index"/) {print "reroute",$3, shard; unassigned=false}}
' > runit
source runit

I'm sure there are better ways to write that script, but great code doesn't typically come in the middle of disasters.

[nik9000]

Pull request is to assign all unassigned primary shards. That would have worked for me in my disaster.

[billzhuang]

it,s great, just want to know when ES can merge @nik9000 PR.

[joestump]

This pull request just saved my butt! Thanks @nik9000. Looking forward to #4285 being pulled in.

[damm]

@nik9000 I forgot to thank you so very much for this pull request. I can't name how many times it's saved me time and effort.

[fawaf]

+1 merge please

[clintongormley]

We want to make progress on this, but the PR seems to provide all the required info. I'll close this issue in favour of the PR: #4285