OpenSearch/docs/reference/ilm/policy-definitions.asciidoc

752 lines
21 KiB
Plaintext

[role="xpack"]
[testenv="basic"]
[[ilm-policy-definition]]
== Policy phases and actions
There are four stages in the index lifecycle, in the order
they are executed.
[options="header"]
|======
| Name | Description
| `hot` | The index is actively being written to
| `warm` | The index is generally not being written to, but is still queried
| `cold` | The index is no longer being updated and is seldom queried. The
information still needs to be searchable, but it's okay if those queries are
slower.
| `delete` | The index is no longer needed and can safely be deleted
|======
Each of these stages is called a "phase". A policy does not need to configure
each phase for an index. For example, one policy may define only the hot
phase and the delete phase, while another may define all four phases.
=== Timing
Indices enter phases based on a phase's `min_age` parameter.
The index will not enter the phase until the index's age is older than that
of the `min_age`. The parameter is configured using a time
duration format (see <<time-units, Time Units>>).
`min_age` defaults to zero seconds `0s` for each phase if not specified.
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"warm": {
"min_age": "1d",
"actions": {
"allocate": {
"number_of_replicas": 1
}
}
},
"delete": {
"min_age": "30d",
"actions": {
"delete": {}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
The Above example configures a policy that moves the index into the warm
phase after one day. Until then, the index is in a waiting state. After
moving into the warm phase, it will wait until 30 days have elapsed before
moving to the delete phase and deleting the index.
`min_age` is usually the time elapsed from the time the index is created. If the
index is rolled over, then `min_age` is the time elapsed from the time the index
is rolled over. The intention here is to execute following phases and actions
relative to when data was written last to a rolled over index.
The previous phase's actions must complete before {ilm} will check `min_age`
and transition into the next phase.
=== Phase Execution
The current phase definition, of an index's policy being executed, is stored
in the index's metadata. The phase and its actions are compiled into a series
of discrete steps that are executed sequentially. Since some {ilm-init} actions
are more complex and involve multiple operations against an index, each of these
operations are done in isolation in a unit called a "step". The
<<ilm-explain-lifecycle,Explain Lifecycle API>> exposes this information to us
to see which step our index is either to execute next, or is currently
executing.
=== Actions
The below list shows the actions which are available in each phase.
NOTE: The order that configured actions are performed in within each phase is
determined by automatically by {ilm-init}, and cannot be changed by changing the
policy definition.
* Hot
- <<ilm-set-priority-action,Set Priority>>
- <<ilm-rollover-action,Rollover>>
- <<ilm-unfollow-action,Unfollow>>
* Warm
- <<ilm-set-priority-action,Set Priority>>
- <<ilm-allocate-action,Allocate>>
- <<ilm-readonly-action,Read-Only>>
- <<ilm-forcemerge-action,Force Merge>>
- <<ilm-shrink-action,Shrink>>
- <<ilm-unfollow-action,Unfollow>>
* Cold
- <<ilm-set-priority-action,Set Priority>>
- <<ilm-allocate-action,Allocate>>
- <<ilm-freeze-action,Freeze>>
- <<ilm-unfollow-action,Unfollow>>
* Delete
- <<ilm-delete-action,Delete>>
[[ilm-allocate-action]]
==== Allocate
Phases allowed: warm, cold.
The Allocate action allows you to specify which nodes are allowed to host the
shards of the index and set the number of replicas.
Behind the scenes, it is modifying the index settings
for shard filtering and/or replica counts. When updating the number of replicas,
configuring allocation rules is optional. When configuring allocation rules,
setting number of replicas is optional. Although this action can be treated as
two separate index settings updates, both can be configured at once.
For more information about how {es} uses replicas for scaling, see
<<scalability>>. See <<shard-allocation-filtering>> for more information about
controlling where Elasticsearch allocates shards of a particular index.
[[ilm-allocate-options]]
.Allocate Options
[options="header"]
|======
| Name | Required | Default | Description
| `number_of_replicas` | no | - | The number of replicas to
assign to the index
| `include` | no | - | assigns an index to nodes
having at least _one_ of the attributes
| `exclude` | no | - | assigns an index to nodes having
_none_ of the attributes
| `require` | no | - | assigns an index to nodes having
_all_ of the attributes
|======
If `number_of_replicas` is not configured, then at least one of `include`,
`exclude`, and `require` is required. An empty Allocate Action with no configuration
is invalid.
===== Example: Change number of replicas
In this example, the index's number of replicas is changed to `2`, while allocation
rules are unchanged.
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"warm": {
"actions": {
"allocate" : {
"number_of_replicas" : 2
}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
===== Example: Assign index to node with specific "box_type" attribute
This example assigns the index to nodes with `box_type` attribute of "hot" or "warm".
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"warm": {
"actions": {
"allocate" : {
"include" : {
"box_type": "hot,warm"
}
}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
===== Example: Assign index to a specific node and update replica settings
This example updates the index to have one replica per shard and be allocated
to nodes with a `box_type` attribute of "cold".
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"warm": {
"actions": {
"allocate" : {
"number_of_replicas": 1,
"require" : {
"box_type": "cold"
}
}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
[[ilm-delete-action]]
==== Delete
Phases allowed: delete.
The Delete Action does just that, it deletes the index.
This action does not have any options associated with it.
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"delete": {
"actions": {
"delete" : { }
}
}
}
}
}
--------------------------------------------------
// CONSOLE
[[ilm-forcemerge-action]]
==== Force Merge
Phases allowed: warm.
NOTE: Index will be be made read-only when this action is run
(see: <<dynamic-index-settings,index.blocks.write>>)
The Force Merge Action <<indices-forcemerge,force merges>> the index into at
most a specific number of <<indices-segments,segments>>.
[[ilm-forcemerge-options]]
.Force Merge Options
[options="header"]
|======
| Name | Required | Default | Description
| `max_num_segments` | yes | - | The number of
segments to merge to.
To fully merge the
index, set it to `1`
|======
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"warm": {
"actions": {
"forcemerge" : {
"max_num_segments": 1
}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
[[ilm-freeze-action]]
==== Freeze
Phases allowed: cold.
This action will <<frozen-indices, freeze>> the index
by calling the <<freeze-index-api, Freeze Index API>>.
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"cold": {
"actions": {
"freeze" : { }
}
}
}
}
}
--------------------------------------------------
// CONSOLE
[IMPORTANT]
================================
Freezing an index will close the index and reopen it within the same API call.
This causes primaries to not be allocated for a short amount of time and
causes the cluster to go red until the primaries are allocated again.
This limitation might be removed in the future.
================================
[[ilm-readonly-action]]
==== Read-Only
Phases allowed: warm.
This action will set the index to be read-only
(see: <<dynamic-index-settings,index.blocks.write>>)
This action does not have any options associated with it.
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"warm": {
"actions": {
"readonly" : { }
}
}
}
}
}
--------------------------------------------------
// CONSOLE
[[ilm-rollover-action]]
==== Rollover
Phases allowed: hot.
[WARNING]
index format must match pattern '^.*-\\d+$', for example (`logs-000001`).
[WARNING]
The managed index must set `index.lifecycle.rollover_alias` as the
alias to rollover. The index must also be the write index for the alias.
[IMPORTANT]
If a policy using the Rollover action is used on a <<ccr-put-follow,follower
index>>, policy execution will wait until the leader index rolls over (or has
<<skipping-rollover, otherwise been marked as complete>>), then convert the
follower index into a regular index as if <<ilm-unfollow-action,the Unfollow
action>> had been used instead of rolling over.
For example, if an index to be managed has an alias `my_data`. The managed
index "my_index" must be the write index for the alias. For more information, read
<<indices-rollover-is-write-index,Write Index Alias Behavior>>.
[source,js]
--------------------------------------------------
PUT my_index
{
"settings": {
"index.lifecycle.name": "my_policy",
"index.lifecycle.rollover_alias": "my_data"
},
"aliases": {
"my_data": {
"is_write_index": true
}
}
}
--------------------------------------------------
// CONSOLE
The Rollover Action rolls an alias over to a new index when the
existing index meets one of the rollover conditions.
[[ilm-rollover-options]]
.Rollover Options
[options="header"]
|======
| Name | Required | Default | Description
| `max_size` | no | - | max primary shard index storage size.
See <<byte-units, Byte Units>>
for formatting
| `max_docs` | no | - | max number of documents an
index is to contain before
rolling over.
| `max_age` | no | - | max time elapsed from index
creation. See
<<time-units, Time Units>>
for formatting
|======
At least one of `max_size`, `max_docs`, `max_age` or any combinations of the
three are required to be specified.
===== Example: Rollover when index is too large
This example rolls the index over when it is at least 100 gigabytes.
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"hot": {
"actions": {
"rollover" : {
"max_size": "100GB"
}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
===== Example: Rollover when index has too many documents
This example rolls the index over when it contains at least
100000000 documents.
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"hot": {
"actions": {
"rollover" : {
"max_docs": 100000000
}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
===== Example: Rollover when index is too old
This example rolls the index over when it has been created at least
7 days ago.
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"hot": {
"actions": {
"rollover" : {
"max_age": "7d"
}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
===== Example: Rollover when index is too old or too large
This example rolls the index over when it has been created at least
7 days ago or it is at least 100 gigabytes. In this case, the index will be
rolled over when any of the conditions is met.
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"hot": {
"actions": {
"rollover" : {
"max_age": "7d",
"max_size": "100GB"
}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
===== Example: Rollover condition stalls phase transition
The Rollover action will only complete once one of its conditions is
met. This means that any proceeding phases will be blocked until Rollover
succeeds.
[source,js]
--------------------------------------------------
PUT /_ilm/policy/rollover_policy
{
"policy": {
"phases": {
"hot": {
"actions": {
"rollover": {
"max_size": "50G"
}
}
},
"delete": {
"min_age": "1d",
"actions": {
"delete": {}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
The above example illustrates a policy which attempts to delete an
index one day after the index has been rolled over. It does not
delete the index one day after it has been created.
[[ilm-set-priority-action]]
==== Set Priority
Phases allowed: hot, warm, cold.
This action sets the <<recovery-prioritization, index priority>> on the index as
soon as the policy enters the hot, warm, or cold phase. Indices with a higher
priority will be recovered before indices with lower priorities following a node
restart. Generally, indexes in the hot phase should have the highest value and
indexes in the cold phase should have the lowest values. For example:
100 for the hot phase, 50 for the warm phase, and 0 for the cold phase.
Indicies that don't set this value have an implicit default priority of 1.
[[ilm-set-priority-options]]
.Set Priority Options
[options="header"]
|======
| Name | Required | Default | Description
| `priority` | yes | - | The priority for the index. Must be 0 or greater.
The value may also be set to null to remove the priority.
|======
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"warm": {
"actions": {
"set_priority" : {
"priority": 50
}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
[[ilm-shrink-action]]
==== Shrink
NOTE: Index will be be made read-only when this action is run
(see: <<dynamic-index-settings,index.blocks.write>>)
[IMPORTANT]
If a policy using the Shrink action is used on a <<ccr-put-follow,follower
index>>, policy execution will wait until the leader index rolls over (or has
<<skipping-rollover, otherwise been marked as complete>>), then convert the
follower index into a regular index as if <<ilm-unfollow-action,the Unfollow
action>> had been used before shrink is applied, as shrink cannot be safely
applied to follower indices.
This action shrinks an existing index into a new index with fewer primary
shards. It calls the <<indices-shrink-index,Shrink API>> to shrink the index.
Since allocating all the primary shards of the index to one node is a
prerequisite, this action will first allocate the primary shards to a valid
node. After shrinking, it will swap aliases pointing to the original index
into the new shrunken index. The new index will also have a new name:
"shrink-<origin-index-name>". So if the original index was called "logs",
then the new index will be named "shrink-logs".
[[ilm-shrink-options]]
.Shrink Options
[options="header"]
|======
| Name | Required | Default | Description
| `number_of_shards` | yes | - | The number of shards
to shrink to. must be
a factor of the number
of shards in the
source index.
|======
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"warm": {
"actions": {
"shrink" : {
"number_of_shards": 1
}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
[[ilm-unfollow-action]]
==== Unfollow
[IMPORTANT]
This action may be used explicitly, as shown below, but this action is also run
before <<ilm-rollover-action,the Rollover action>> and <<ilm-shrink-action,the
Shrink action>> as described in the documentation for those actions.
This action turns a {ref}/ccr-apis.html[ccr] follower index
into a regular index. This can be desired when moving follower
indices into the next phase. Also certain actions like shrink
and rollover can then be performed safely on follower indices.
This action will wait until is it safe to convert a follower index into a
regular index. In particular, the following conditions must be met:
* The leader index must have `index.lifecycle.indexing_complete` set to `true`.
This happens automatically if the leader index is rolled over using
<<ilm-rollover-action,the Rollover action>>, or may be set manually using
the <<indices-update-settings,Index Settings API>>.
* All operations performed on the leader index must have been replicated to the
follower index. This ensures that no operations will be lost when the index is
converted into a regular index.
If the unfollow action encounters a follower index then
the following operations will be performed on it:
* Pauses indexing following for the follower index.
* Closes the follower index.
* Unfollows the follower index.
* Opens the follower index (which is at this point is a regular index).
The unfollow action does not have any options and
if it encounters a non follower index, then the
unfollow action leaves that index untouched and
lets the next action operate on this index.
[source,js]
--------------------------------------------------
PUT _ilm/policy/my_policy
{
"policy": {
"phases": {
"hot": {
"actions": {
"unfollow" : {}
}
}
}
}
}
--------------------------------------------------
// CONSOLE
=== Full Policy
With all of these actions, we can support complex management strategies for our
indices. This policy will define an index that will start in the hot phase,
rolling over every 50 GB or 7 days. After 30 days it enters the warm phase
and increases the replicas to 2, force merges and shrinks. After 60 days
it enters the cold phase and allocates to "cold" nodes, and after 90 days the
index is deleted.
[source,js]
--------------------------------------------------
PUT _ilm/policy/full_policy
{
"policy": {
"phases": {
"hot": {
"actions": {
"rollover": {
"max_age": "7d",
"max_size": "50G"
}
}
},
"warm": {
"min_age": "30d",
"actions": {
"forcemerge": {
"max_num_segments": 1
},
"shrink": {
"number_of_shards": 1
},
"allocate": {
"number_of_replicas": 2
}
}
},
"cold": {
"min_age": "60d",
"actions": {
"allocate": {
"require": {
"type": "cold"
}
}
}
},
"delete": {
"min_age": "90d",
"actions": {
"delete": {}
}
}
}
}
}
--------------------------------------------------
// CONSOLE