Marvel Problem with some dashboards

Ηello guys,

i had my marvel page working fine some hours ago, but now i have a problem. The problem is that Overview page as well as Shard Allocation page cannot be loaded. In overview page i take empty diagramms and in shard allocation i get "Error The connection to Elasticsearch returned an error. Check your Elasticsearch instance."

The thing is that i can load cluster pulse Node and Index statistics pages and of course elasticsearch is working properly. What the problem may is?

Is there anything in your Elasticsearch logs that might point to a problem?

Yes, each time i refresh the page the following log occurs:

ubuntu@logstashganglia:/var/log/elasticsearch$ tail cslabES.log
at java.util.concurrent.ThreadPoolExecutor.runWorker(
at java.util.concurrent.ThreadPoolExecutor$
Caused by: [.marvel-2015.05.27][0]: from[-1],size[1]: Parse Failure [No mapping found for [@timestamp] in order to sort on]
... 9 more
[2015-05-27 09:25:41,551][DEBUG][ ] [MasterNode] All shards failed for phase: [query_fetch]

I think the problem occured when the master node changed from one physical node to onather one. I "played" a lot by changing who would be the master node and now overview works fine. However i still cannot load shard alocation page. The log is the same:
[2015-05-27 11:04:19,253][DEBUG][ ] [ES8] All shards failed for phase: [query_fetch]

It might be obvious but, be sure that Marvel plugin is installed on each nodes (or ES instances).