elasticsearch/docs/reference/how-to/size-your-shards.asciidoc

[[size-your-shards]]
== Size your shards

Each index in {es} is divided into one or more shards, each of which may be
replicated across multiple nodes to protect against hardware failures. If you
are using <<data-streams>> then each data stream is backed by a sequence of
indices. There is a limit to the amount of data you can store on a single node
so you can increase the capacity of your cluster by adding nodes and increasing
the number of indices and shards to match. However, each index and shard has
some overhead and if you divide your data across too many shards then the
overhead can become overwhelming. A cluster with too many indices or shards is
said to suffer from _oversharding_. An oversharded cluster will be less
efficient at responding to searches and in extreme cases it may even become
unstable.

[discrete]
[[create-a-sharding-strategy]]
=== Create a sharding strategy

The best way to prevent oversharding and other shard-related issues is to
create a sharding strategy. A sharding strategy helps you determine and
maintain the optimal number of shards for your cluster while limiting the size
of those shards.

Unfortunately, there is no one-size-fits-all sharding strategy. A strategy that
works in one environment may not scale in another. A good sharding strategy
must account for your infrastructure, use case, and performance expectations.

The best way to create a sharding strategy is to benchmark your production data
on production hardware using the same queries and indexing loads you'd see in
production. For our recommended methodology, watch the
https://www.elastic.co/elasticon/conf/2016/sf/quantitative-cluster-sizing[quantitative
cluster sizing video]. As you test different shard configurations, use {kib}'s
{kibana-ref}/elasticsearch-metrics.html[{es} monitoring tools] to track your
cluster's stability and performance.

The following sections provide some reminders and guidelines you should
consider when designing your sharding strategy. If your cluster is already
oversharded, see <<reduce-cluster-shard-count>>.

[discrete]
[[shard-sizing-considerations]]
=== Sizing considerations

Keep the following things in mind when building your sharding strategy.

[discrete]
[[single-thread-per-shard]]
==== Searches run on a single thread per shard

Most searches hit multiple shards. Each shard runs the search on a single
CPU thread. While a shard can run multiple concurrent searches, searches across a
large number of shards can deplete a node's <<modules-threadpool,search
thread pool>>. This can result in low throughput and slow search speeds.

[discrete]
[[each-shard-has-overhead]]
==== Each index, shard, segment and field has overhead

Every index and every shard requires some memory and CPU resources. In most
cases, a small set of large shards uses fewer resources than many small shards.

Segments play a big role in a shard's resource usage. Most shards contain
several segments, which store its index data. {es} keeps some segment metadata
in heap memory so it can be quickly retrieved for searches. As a shard grows,
its segments are <<index-modules-merge,merged>> into fewer, larger segments.
This decreases the number of segments, which means less metadata is kept in
heap memory.

Every mapped field also carries some overhead in terms of memory usage and disk
space. By default {es} will automatically create a mapping for every field in
every document it indexes, but you can switch off this behaviour to
<<explicit-mapping,take control of your mappings>>.

Moreover every segment requires a small amount of heap memory for each mapped
field. This per-segment-per-field heap overhead includes a copy of the field
name, encoded using ISO-8859-1 if applicable or UTF-16 otherwise. Usually this
is not noticeable, but you may need to account for this overhead if your shards
have high segment counts and the corresponding mappings contain high field
counts and/or very long field names.

[discrete]
[[shard-auto-balance]]
==== {es} automatically balances shards within a data tier

A cluster's nodes are grouped into <<data-tiers,data tiers>>. Within each tier,
{es} attempts to spread an index's shards across as many nodes as possible. When
you add a new node or a node fails, {es} automatically rebalances the index's
shards across the tier's remaining nodes.

[discrete]
[[shard-size-best-practices]]
=== Best practices

Where applicable, use the following best practices as starting points for your
sharding strategy.

[discrete]
[[delete-indices-not-documents]]
==== Delete indices, not documents

Deleted documents aren't immediately removed from {es}'s file system.
Instead, {es} marks the document as deleted on each related shard. The marked
document will continue to use resources until it's removed during a periodic
<<index-modules-merge,segment merge>>.

When possible, delete entire indices instead. {es} can immediately remove
deleted indices directly from the file system and free up resources.

[discrete]
[[use-ds-ilm-for-time-series]]
==== Use data streams and {ilm-init} for time series data

<<data-streams,Data streams>> let you store time series data across multiple,
time-based backing indices. You can use <<index-lifecycle-management,{ilm}
({ilm-init})>> to automatically manage these backing indices.

One advantage of this setup is
<<getting-started-index-lifecycle-management,automatic rollover>>, which creates
a new write index when the current one meets a defined `max_primary_shard_size`,
`max_age`, `max_docs`, or `max_size` threshold. When an index is no longer
needed, you can use {ilm-init} to automatically delete it and free up resources.

{ilm-init} also makes it easy to change your sharding strategy over time:

* *Want to decrease the shard count for new indices?* +
Change the <<index-number-of-shards,`index.number_of_shards`>> setting in the
data stream's <<data-streams-change-mappings-and-settings,matching index
template>>.

* *Want larger shards or fewer backing indices?* +
Increase your {ilm-init} policy's <<ilm-rollover,rollover threshold>>.

* *Need indices that span shorter intervals?* +
Offset the increased shard count by deleting older indices sooner. You can do
this by lowering the `min_age` threshold for your policy's
<<ilm-index-lifecycle,delete phase>>.

Every new backing index is an opportunity to further tune your strategy.

[discrete]
[[shard-size-recommendation]]
==== Aim for shard sizes between 10GB and 50GB

Larger shards take longer to recover after a failure. When a node fails, {es}
rebalances the node's shards across the data tier's remaining nodes. This
recovery process typically involves copying the shard contents across the
network, so a 100GB shard will take twice as long to recover than a 50GB shard.
In contrast, small shards carry proportionally more overhead and are less
efficient to search. Searching fifty 1GB shards will take substantially more
resources than searching a single 50GB shard containing the same data.

There are no hard limits on shard size, but experience shows that shards
between 10GB and 50GB typically work well for logs and time series data. You
may be able to use larger shards depending on your network and use case.
Smaller shards may be appropriate for
{enterprise-search-ref}/index.html[Enterprise Search] and similar use cases.

If you use {ilm-init}, set the <<ilm-rollover,rollover action>>'s
`max_primary_shard_size` threshold to `50gb` to avoid shards larger than 50GB.

To see the current size of your shards, use the <<cat-shards,cat shards API>>.

[source,console]
----
GET _cat/shards?v=true&h=index,prirep,shard,store&s=prirep,store&bytes=gb
----
// TEST[setup:my_index]

The `pri.store.size` value shows the combined size of all primary shards for
the index.

[source,txt]
----
index                                 prirep shard store
.ds-my-data-stream-2099.05.06-000001  p      0      50gb
...
----
// TESTRESPONSE[non_json]
// TESTRESPONSE[s/\.ds-my-data-stream-2099\.05\.06-000001/my-index-000001/]
// TESTRESPONSE[s/50gb/.*/]

[discrete]
[[shard-count-recommendation]]
==== Master-eligible nodes should have at least 1GB of heap per 3000 indices

The number of indices a master node can manage is proportional to its heap
size. The exact amount of heap memory needed for each index depends on various
factors such as the size of the mapping and the number of shards per index.

As a general rule of thumb, you should have fewer than 3000 indices per GB of
heap on master nodes. For example, if your cluster has dedicated master nodes
with 4GB of heap each then you should have fewer than 12000 indices. If your
master nodes are not dedicated master nodes then the same sizing guidance
applies: you should reserve at least 1GB of heap on each master-eligible node
for every 3000 indices in your cluster.

Note that this rule defines the absolute maximum number of indices that a
master node can manage, but does not guarantee the performance of searches or
indexing involving this many indices. You must also ensure that your data nodes
have adequate resources for your workload and that your overall sharding
strategy meets all your performance requirements. See also
<<single-thread-per-shard>> and <<each-shard-has-overhead>>.

To check the configured size of each node's heap, use the <<cat-nodes,cat nodes
API>>.

[source,console]
----
GET _cat/nodes?v=true&h=heap.max
----
// TEST[setup:my_index]

You can use the <<cat-shards,cat shards API>> to check the number of shards per
node.

[source,console]
----
GET _cat/shards?v=true
----
// TEST[setup:my_index]

[discrete]
[[field-count-recommendation]]
==== Allow enough heap for field mappers and overheads

Mapped fields consume some heap memory on each node, and require extra
heap on data nodes.
Ensure each node has enough heap for mappings, and also allow
extra space for overheads associated with its workload. The following sections
show how to determine these heap requirements.

[discrete]
===== Mapping metadata in the cluster state

Each node in the cluster has a copy of the <<cluster-state-api-desc,cluster state>>.
The cluster state includes information about the field mappings for
each index. This information has heap overhead. You can use the
<<cluster-stats,Cluster stats API>> to get the heap overhead of the total size of
all mappings after deduplication and compression.

[source,console]
----
GET _cluster/stats?human&filter_path=indices.mappings.total_deduplicated_mapping_size*
----
// TEST[setup:node]

This will show you information like in this example output:

[source,console-result]
----
{
  "indices": {
    "mappings": {
      "total_deduplicated_mapping_size": "1gb",
      "total_deduplicated_mapping_size_in_bytes": 1073741824
    }
  }
}
----
// TESTRESPONSE[s/"total_deduplicated_mapping_size": "1gb"/"total_deduplicated_mapping_size": $body.$_path/]
// TESTRESPONSE[s/"total_deduplicated_mapping_size_in_bytes": 1073741824/"total_deduplicated_mapping_size_in_bytes": $body.$_path/]

[discrete]
===== Retrieving heap size and field mapper overheads

You can use the <<cluster-nodes-stats,Nodes stats API>> to get two relevant metrics
for each node:

* The size of the heap on each node.

* Any additional estimated heap overhead for the fields per node. This is specific to
data nodes, where apart from the cluster state field information mentioned above,
there is additional heap overhead for each mapped field of an index held by the data
node. For nodes which are not data nodes, this field may be zero.

[source,console]
----
GET _nodes/stats?human&filter_path=nodes.*.name,nodes.*.indices.mappings.total_estimated_overhead*,nodes.*.jvm.mem.heap_max*
----
// TEST[setup:node]

For each node, this will show you information like in this example output:

[source,console-result]
----
{
  "nodes": {
    "USpTGYaBSIKbgSUJR2Z9lg": {
      "name": "node-0",
      "indices": {
        "mappings": {
          "total_estimated_overhead": "1gb",
          "total_estimated_overhead_in_bytes": 1073741824
        }
      },
      "jvm": {
        "mem": {
          "heap_max": "4gb",
          "heap_max_in_bytes": 4294967296
        }
      }
    }
  }
}
----
// TESTRESPONSE[s/"USpTGYaBSIKbgSUJR2Z9lg"/\$node_name/]
// TESTRESPONSE[s/"name": "node-0"/"name": $body.$_path/]
// TESTRESPONSE[s/"total_estimated_overhead": "1gb"/"total_estimated_overhead": $body.$_path/]
// TESTRESPONSE[s/"total_estimated_overhead_in_bytes": 1073741824/"total_estimated_overhead_in_bytes": $body.$_path/]
// TESTRESPONSE[s/"heap_max": "4gb"/"heap_max": $body.$_path/]
// TESTRESPONSE[s/"heap_max_in_bytes": 4294967296/"heap_max_in_bytes": $body.$_path/]

[discrete]
===== Consider additional heap overheads

Apart from the two field overhead metrics above, you must additionally allow
enough heap for {es}'s baseline usage as well as your workload such as indexing,
searches and aggregations. 0.5GB of extra heap will suffice for many reasonable
workloads, and you may need even less if your workload is very light while heavy
workloads may require more.

[discrete]
===== Example

As an example, consider the outputs above for a data node. The heap of the node
will need at least:

* 1 GB for the cluster state field information.

* 1 GB for the additional estimated heap overhead for the fields of the data node.

* 0.5 GB of extra heap for other overheads.

Since the node has a 4GB heap max size in the example, it is thus sufficient
for the total required heap of 2.5GB.

If the heap max size for a node is not sufficient, consider
<<avoid-unnecessary-fields,avoiding unnecessary fields>>,
or scaling up the cluster, or redistributing index shards.

Note that the above rules do not necessarily guarantee the performance of
searches or indexing involving a very high number of indices. You must also
ensure that your data nodes have adequate resources for your workload and
that your overall sharding strategy meets all your performance requirements.
See also <<single-thread-per-shard>> and <<each-shard-has-overhead>>.

[discrete]
[[avoid-node-hotspots]]
==== Avoid node hotspots

If too many shards are allocated to a specific node, the node can become a
hotspot. For example, if a single node contains too many shards for an index
with a high indexing volume, the node is likely to have issues.

To prevent hotspots, use the
<<total-shards-per-node,`index.routing.allocation.total_shards_per_node`>> index
setting to explicitly limit the number of shards on a single node. You can
configure `index.routing.allocation.total_shards_per_node` using the
<<indices-update-settings,update index settings API>>.

[source,console]
--------------------------------------------------
PUT my-index-000001/_settings
{
  "index" : {
    "routing.allocation.total_shards_per_node" : 5
  }
}
--------------------------------------------------
// TEST[setup:my_index]

[discrete]
[[avoid-unnecessary-fields]]
==== Avoid unnecessary mapped fields

By default {es} <<dynamic-mapping,automatically creates a mapping>> for every
field in every document it indexes. Every mapped field corresponds to some data
structures on disk which are needed for efficient search, retrieval, and
aggregations on this field. Details about each mapped field are also held in
memory. In many cases this overhead is unnecessary because a field is not used
in any searches or aggregations. Use <<explicit-mapping>> instead of dynamic
mapping to avoid creating fields that are never used. If a collection of fields
are typically used together, consider using <<copy-to>> to consolidate them at
index time. If a field is only rarely used, it may be better to make it a
<<runtime,Runtime field>> instead.

You can get information about which fields are being used with the
<<field-usage-stats>> API, and you can analyze the disk usage of mapped fields
using the <<indices-disk-usage>> API. Note however that unnecessary mapped
fields also carry some memory overhead as well as their disk usage.

[discrete]
[[reduce-cluster-shard-count]]
=== Reduce a cluster's shard count

If your cluster is already oversharded, you can use one or more of the following
methods to reduce its shard count.

[discrete]
[[create-indices-that-cover-longer-time-periods]]
==== Create indices that cover longer time periods

If you use {ilm-init} and your retention policy allows it, avoid using a
`max_age` threshold for the rollover action. Instead, use
`max_primary_shard_size` to avoid creating empty indices or many small shards.

If your retention policy requires a `max_age` threshold, increase it to create
indices that cover longer time intervals. For example, instead of creating daily
indices, you can create indices on a weekly or monthly basis.

[discrete]
[[delete-empty-indices]]
==== Delete empty or unneeded indices

If you're using {ilm-init} and roll over indices based on a `max_age` threshold,
you can inadvertently create indices with no documents. These empty indices
provide no benefit but still consume resources.

You can find these empty indices using the <<cat-count,cat count API>>.

[source,console]
----
GET _cat/count/my-index-000001?v=true
----
// TEST[setup:my_index]

Once you have a list of empty indices, you can delete them using the
<<indices-delete-index,delete index API>>. You can also delete any other
unneeded indices.

[source,console]
----
DELETE my-index-000001
----
// TEST[setup:my_index]

[discrete]
[[force-merge-during-off-peak-hours]]
==== Force merge during off-peak hours

If you no longer write to an index, you can use the <<indices-forcemerge,force
merge API>> to <<index-modules-merge,merge>> smaller segments into larger ones.
This can reduce shard overhead and improve search speeds. However, force merges
are resource-intensive. If possible, run the force merge during off-peak hours.

[source,console]
----
POST my-index-000001/_forcemerge
----
// TEST[setup:my_index]

[discrete]
[[shrink-existing-index-to-fewer-shards]]
==== Shrink an existing index to fewer shards

If you no longer write to an index, you can use the
<<indices-shrink-index,shrink index API>> to reduce its shard count.

{ilm-init} also has a <<ilm-shrink,shrink action>> for indices in the
warm phase.

[discrete]
[[combine-smaller-indices]]
==== Combine smaller indices

You can also use the <<docs-reindex,reindex API>> to combine indices
with similar mappings into a single large index. For time series data, you could
reindex indices for short time periods into a new index covering a
longer period. For example, you could reindex daily indices from October with a
shared index pattern, such as `my-index-2099.10.11`, into a monthly
`my-index-2099.10` index. After the reindex, delete the smaller indices.

[source,console]
----
POST _reindex
{
  "source": {
    "index": "my-index-2099.10.*"
  },
  "dest": {
    "index": "my-index-2099.10"
  }
}
----

[discrete]
[[troubleshoot-shard-related-errors]]
=== Troubleshoot shard-related errors

Here’s how to resolve common shard-related errors.

[discrete]
==== this action would add [x] total shards, but this cluster currently has [y]/[z] maximum shards open;

The <<cluster-max-shards-per-node,`cluster.max_shards_per_node`>> cluster
setting limits the maximum number of open shards for a cluster. This error
indicates an action would exceed this limit.

If you're confident your changes won't destabilize the cluster, you can
temporarily increase the limit using the <<cluster-update-settings,cluster
update settings API>> and retry the action.

[source,console]
----
PUT _cluster/settings
{
  "persistent" : {
    "cluster.max_shards_per_node": 1200
  }
}
----

This increase should only be temporary. As a long-term solution, we recommend
you add nodes to the oversharded data tier or
<<reduce-cluster-shard-count,reduce your cluster's shard count>>. To get a
cluster's current shard count after making changes, use the
<<cluster-stats,cluster stats API>>.

[source,console]
----
GET _cluster/stats?filter_path=indices.shards.total
----

When a long-term solution is in place, we recommend you reset the
`cluster.max_shards_per_node` limit.

[source,console]
----
PUT _cluster/settings
{
  "persistent" : {
    "cluster.max_shards_per_node": null
  }
}
----