[Query] create index fails with an error There is no available index service even when indexer is active and running

Steps:
1. Create and configure cluster
2. Create buckets and load around 300k documents
3. Create primary index

Index creation fails with an error:
"requestID": "04a6c913-cca3-4270-b892-7b8c2c89f3ac",
"signature": null,
"results": [
],
"errors": [

{ "code": 5000, "msg": "GSI CreatePrimaryIndex() - cause: Fails to create index. There is no available index service that can process this request at this time. Index Service can be in bootstrap, recovery, or non-reachable. Please retry the operation at a later time." }

],
"status": "errors",
"metrics":

{ "elapsedTime": "10.008678132s", "executionTime": "10.00850264s", "resultCount": 0, "resultSize": 0, "errorCount": 1 }

}

This is the regression from 4.5.1-2806 to 4.5.1-2816/2817.
The issue is observed in simple tests on Windows machines as well.

Testrunner command:
./testrunner -i /tmp/centos_upgrade2i.ini use_gsi_for_primary=True,defer_build=True,initial_index_number=2,get-cbcollect-info=True,get-coredump=True -t 2i.memdb_oom_2i.SecondaryIndexMemdbOomTests.test_oom_flush_bucket,gsi_type=memory_optimized,nodes_init=4,services_init=kv:n1ql-kv-index-index,doc_ops=True,standard_buckets=2,default_buckets=1,initial_index_number=4,scan_consistency=request_plus

Testrunner Logs: http://qa.sc.couchbase.com/job/cen7-2i-set6-job1-moi-oom/178/consoleFull

Cluster logs:
https://s3.amazonaws.com/bugdb/jira/indexer_unavailable/collectinfo-2016-08-18T120642-ns_1%40172.23.105.90.zip
https://s3.amazonaws.com/bugdb/jira/indexer_unavailable/collectinfo-2016-08-18T120642-ns_1%40172.23.106.85.zip
https://s3.amazonaws.com/bugdb/jira/indexer_unavailable/collectinfo-2016-08-18T120642-ns_1%40172.23.108.163.zip
https://s3.amazonaws.com/bugdb/jira/indexer_unavailable/collectinfo-2016-08-18T120642-ns_1%40172.23.108.165.zip