Search Node Only

(Pablo Pinargote) #1

Hi, I am very new to elasticsearch, at my company we are testing the data base and we are trying to create a cluster with 2 nodes, one master node and one only search node.

We already had the two nodes up and running, at the search node config we spcified the values:

node.master: false false

And the cluster shows the two nodes already, but, when we check the node diagnostics using HQ we see that the search node has no activity, is something missing on the configuration file?

Thanks in advance.
PD: We do not need a replica node, the data we are storing is volatile so it is not importan to us to have a replication schema, we just one to set up one master and data node, and one search only node.

This is our cluster diagnostics:

Node Name: master search

Store Size: 48.5GB 0.0

Documents: 129,229,804 0

Documents Deleted: 0% 0%
Merge Size: 16.6TB 0.0
Merge Time: 18:42:38 00:00:00
Merge Rate: 6.6 MB/s 0 MB/s
File Descriptors: 14,524 1,746
Index Activity
Indexing - Index: 2.19ms 0ms
Indexing - Delete: 0.31ms 0ms
Search - Query: 285.44ms 0ms
Search - Fetch: 56.36ms 0ms
Get - Total: 0.1ms 0ms
Get - Exists: 0.1ms 0ms
Get - Missing: 0ms 0ms
Refresh: 42.41ms 0ms
Flush: 1202.28ms 0ms
Cache Activity
Field Size: 1.9GB 0.0
Field Evictions: 0 0
Filter Cache Size: 114.6MB 0.0
Filter Evictions: 0 per query 0 per query
ID Cache Size:
% ID Cache: 0% 0%
Total Memory: 32 gb 24 gb
Heap Size: 8.9 gb 4.9 gb
Heap % of RAM: 27.7% 20.4%
% Heap Used: 65.3% 12.8%
GC MarkSweep Frequency: 0 s 0 s
GC MarkSweep Duration: 0ms 0ms
GC ParNew Frequency: 0 s 0 s
GC ParNew Duration: 0ms 0ms
G1 GC Young Generation Freq: 0 s 0 s
G1 GC Young Generation Duration: 0ms 0ms
G1 GC Old Generation Freq: 0 s 0 s
G1 GC Old Generation Duration: 0ms 0ms
Swap Space: 12,775.4648 mb 14,131.8945 mb
HTTP Connection Rate: 0 /second 0 /second
How this Field is Calculated
High values indicate slow I/O.
Keys: stats.indices.flush.total_time_in_millis /
Value: 54094103 / 44993 = 1202.2781988309293
1202.2781988309293 <= 750 = Pass!
1202.2781988309293 <= 1500 = Warning!

(Jason Tedor) #2

For the node that you call a "search node" (sometimes you'll see this referred to as a "client node", or a "smart load balancer") to have any activity, you have to issue requests against it. This node will then do the work of reaching out to the master/data node that you have to coordinate queries.

Your mileage may vary, but it seems very unlikely to me that you need a client node with a single data/master node. The client node can be useful in cases when coordinating requests across a larger cluster, but in the case of one data/master node there isn't much coordination work to be offloaded.

As an aside, I note that you're using the G1 garbage collector; note that at this time use of the G1 garbage collector is discouraged with Elasticsearch.

(Mark Walkom) #3

Not just discouraged, unsupported-and-you-will-catch-fire.

(system) #4