Answer by Rajesh J. ·
Hi Pierre
I couldn't see any transport call which was happening over either TCP or HTTP. I believe the search request you executed in the cluster was fulfilled within "C". I might be wrong... can you do couple of things. Check with Marvel how many nodes every involved for that search request. You might have multiple nodes in the cluster, but for the purepath you have attached, seems they are not involved. Second, can you enable HTTP transport for node to node communication. http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-transport.html
I can reach out to you directly over email to get more details and follow-up webex.
Thanks,
Rajesh
Answer by Pierre D. ·
thanks Rajesh,
in fact in can't see neither in transaction flow nor in purepath detail the time spent by the other node.
Here a sample::
An elasticsearch node "C" which is the client, performs some actions and send a request to another elasticsearch node "E".
I have an dynatrace agent on "C" and configured a start purepath on servlet, intercepting the REST calls, and I have no agent on node "E"..When node C calls node E, I don't see call to E as a Domain less n transaction flow. So I can't have the time division spent in node E.
Is there a way to see the node E as a DomainLess and by this get the time spent by the node E ?
Pierre
Answer by Rajesh J. ·
Pierre
Have you looked at Marvel (http://www.elasticsearch.org/guide/en/marvel/current/index.html)
It gives you a good insight in the cluster. My guess is that if you have to instrument, the client (which could be using REST) is adequate, if you can measure the round trip of the distribute search from the client.
And you can use Marvel (Sense) API's to bring the data into dynaTrace using plugin's
Rajesh
Answer by Pierre D. ·
Hi Reinhard,
I don't know the implementation used by ElasticSearch, I wished Dynatrace can handle this framework (which is well known).
But when I really think about it, I ask myself about the pertinence of this.
Monitoring tools are available for ElasticSearch, and having pure path of Electicsearch implementation may be useless (except to get the time spent in an elacticsearch node).
Is there a mean to have, in a purepath, the time spent by by the client to wait the process done by a node (having in transaction flow an external service call for instance) ?
Answer by Pierre D. ·
I mean without coding a specific part using ADK to handle the correlation id.
Pierre
Pierre,
that completely depends on the specific application protocol beeing used. If it's webservices, RMI, or http based protocols your chances are good to get a connected PurePath.
Reinhard
JANUARY 15, 3:00 PM GMT / 10:00 AM ET