Marking shard [index_name][1] as inactive (inactive_time[])

[Balamurali]

Hi all,

we are using elasticsearch 1.3.7 version and cluster with five data cum master nodes.
since last week , we facing frequently very slowness in cluster. During analysis, we found below debug messages are printed many times in logs.

[2017-02-01 05:44:24,775][DEBUG][indices.memory ] [node_name] marking shard [index_name][1] as inactive (inactive_time[30m]) indexing wise, setting size to [500kb]
[2017-02-01 05:44:24,775][DEBUG][index.engine.internal ] [node_name] [index_name][1] updating index_buffer_size from [401mb] to (inactive) [500kb]
[2017-02-01 05:44:24,800][DEBUG][indices.memory ] [node_name] recalculating shard indexing buffer (reason=[[BECAME_INACTIVE]]), total is [1.1gb] with [2] active shards, each shard set to indexing=[512mb], translog=[64kb]
[2017-02-01 05:44:24,800][DEBUG][index.engine.internal ] [node_name] [index_name][3] updating index_buffer_size from [401mb] to [512mb]
[2017-02-01 05:44:24,800][DEBUG][index.engine.internal ] [node_name] [index_name][4] updating index_buffer_size from [401mb] to [512mb]
[2017-02-01 05:45:07,887][DEBUG][script ] [node_name] notifying script services of script removal due to: [SIZE]
[2017-02-01 05:45:24,800][DEBUG][indices.memory ] [node_name] marking shard [index_name][1] as active indexing wise
[2017-02-01 05:45:24,801][DEBUG][indices.memory ] [node_name] recalculating shard indexing buffer (reason=[[BECAME_ACTIVE]]), total is [1.1gb] with [3] active shards, each shard set to indexing=[401mb], translog=[64kb]

whether above messages are reason for slowness?? what is these messages means???

any one please help us!!!!!

[Herve_Bry (Hervé Bry)]

It means that no data has been indexed in the said shard for more than 30min, so the shard is marked as inactive and its indexation buffers are reduced to free up memory. When new data is indexed, the shard will be marked as active again, and the buffers will grow up to achieve better performace.

With ES 1.x, the buffer can take 30s to grow after the indexing has started again, thus leading to an uncontrolled growth of segments and consequently a lot (a sh*tload might I even say:)) of merges. This has been fixed in ES 2+.

As a side note, unless you have very good reason to stick with ES 1.3.7, I strongly suggest you to upgrade at the very least to ES 1.7, as a lot of bugs and performance issues have been fixed since.