OpenSearch/docs/reference/rollup/apis/rollup-job-config.asciidoc

281 lines
13 KiB
Plaintext

[role="xpack"]
[testenv="basic"]
[[rollup-job-config]]
=== Rollup job configuration
experimental[]
The Rollup Job Configuration contains all the details about how the rollup job should run, when it indexes documents,
and what future queries will be able to execute against the rollup index.
There are three main sections to the Job Configuration; the logistical details about the job (cron schedule, etc), what fields
should be grouped on, and what metrics to collect for each group.
A full job configuration might look like this:
[source,js]
--------------------------------------------------
PUT _rollup/job/sensor
{
"index_pattern": "sensor-*",
"rollup_index": "sensor_rollup",
"cron": "*/30 * * * * ?",
"page_size" :1000,
"groups" : {
"date_histogram": {
"field": "timestamp",
"interval": "60m",
"delay": "7d"
},
"terms": {
"fields": ["hostname", "datacenter"]
},
"histogram": {
"fields": ["load", "net_in", "net_out"],
"interval": 5
}
},
"metrics": [
{
"field": "temperature",
"metrics": ["min", "max", "sum"]
},
{
"field": "voltage",
"metrics": ["avg"]
}
]
}
--------------------------------------------------
// CONSOLE
// TEST[setup:sensor_index]
==== Logistical Details
In the above example, there are several pieces of logistical configuration for the job itself.
`{job_id}` (required)::
(string) In the endpoint URL, you specify the name of the job (`sensor` in the above example). This can be any alphanumeric string,
and uniquely identifies the data that is associated with the rollup job. The ID is persistent, in that it is stored with the rolled
up data. So if you create a job, let it run for a while, then delete the job... the data that the job rolled up will still be
associated with this job ID. You will be unable to create a new job with the same ID, as that could lead to problems with mismatched
job configurations
`index_pattern` (required)::
(string) The index, or index pattern, that you wish to rollup. Supports wildcard-style patterns (`logstash-*`). The job will
attempt to rollup the entire index or index-pattern. Once the "backfill" is finished, it will periodically (as defined by the cron)
look for new data and roll that up too.
`rollup_index` (required)::
(string) The index that you wish to store rollup results into. All the rollup data that is generated by the job will be
stored in this index. When searching the rollup data, this index will be used in the <<rollup-search,Rollup Search>> endpoint's URL.
The rollup index can be shared with other rollup jobs. The data is stored so that it doesn't interfere with unrelated jobs.
`cron` (required)::
(string) A cron string which defines when the rollup job should be executed. The cron string defines an interval of when to run
the job's indexer. When the interval triggers, the indexer will attempt to rollup the data in the index pattern. The cron pattern
is unrelated to the time interval of the data being rolled up. For example, you may wish to create hourly rollups of your document (as
defined in the <<rollup-groups-config,grouping configuration>>) but to only run the indexer on a daily basis at midnight, as defined by the cron.
The cron pattern is defined just like Watcher's Cron Schedule.
`page_size` (required)::
(int) The number of bucket results that should be processed on each iteration of the rollup indexer. A larger value
will tend to execute faster, but will require more memory during processing. This has no effect on how the data is rolled up, it is
merely used for tweaking the speed/memory cost of the indexer.
[NOTE]
The `index_pattern` cannot be a pattern that would also match the destination `rollup_index`. E.g. the pattern
`"foo-*"` would match the rollup index `"foo-rollup"`. This causes problems because the rollup job would attempt
to rollup it's own data at runtime. If you attempt to configure a pattern that matches the `rollup_index`, an exception
will be thrown to prevent this behavior.
[[rollup-groups-config]]
==== Grouping Config
The `groups` section of the configuration is where you decide which fields should be grouped on, and with what aggregations. These
fields will then be available later for aggregating into buckets. For example, this configuration:
[source,js]
--------------------------------------------------
"groups" : {
"date_histogram": {
"field": "timestamp",
"interval": "60m",
"delay": "7d"
},
"terms": {
"fields": ["hostname", "datacenter"]
},
"histogram": {
"fields": ["load", "net_in", "net_out"],
"interval": 5
}
}
--------------------------------------------------
// NOTCONSOLE
Allows `date_histogram`'s to be used on the `"timestamp"` field, `terms` aggregations to be used on the `"hostname"` and `"datacenter"`
fields, and `histograms` to be used on any of `"load"`, `"net_in"`, `"net_out"` fields.
Importantly, these aggs/fields can be used in any combination. Think of the `groups` configuration as defining a set of tools that can
later be used in aggregations to partition the data. Unlike raw data, we have to think ahead to which fields and aggregations might be used.
But Rollups provide enough flexibility that you simply need to determine _which_ fields are needed, not _in what order_ they are needed.
There are three types of groupings currently available:
===== Date Histogram
A `date_histogram` group aggregates a `date` field into time-based buckets. The `date_histogram` group is *mandatory* -- you currently
cannot rollup documents without a timestamp and a `date_histogram` group.
The `date_histogram` group has several parameters:
`field` (required)::
The date field that is to be rolled up.
`interval` (required)::
The interval of time buckets to be generated when rolling up. E.g. `"60m"` will produce 60 minute (hourly) rollups. This follows standard time formatting
syntax as used elsewhere in Elasticsearch. The `interval` defines the _minimum_ interval that can be aggregated only. If hourly (`"60m"`)
intervals are configured, <<rollup-search,Rollup Search>> can execute aggregations with 60m or greater (weekly, monthly, etc) intervals.
So define the interval as the smallest unit that you wish to later query.
Note: smaller, more granular intervals take up proportionally more space.
`delay`::
How long to wait before rolling up new documents. By default, the indexer attempts to roll up all data that is available. However, it
is not uncommon for data to arrive out of order, sometimes even a few days late. The indexer is unable to deal with data that arrives
after a time-span has been rolled up (e.g. there is no provision to update already-existing rollups).
Instead, you should specify a `delay` that matches the longest period of time you expect out-of-order data to arrive. E.g. a `delay` of
`"1d"` will instruct the indexer to roll up documents up to `"now - 1d"`, which provides a day of buffer time for out-of-order documents
to arrive.
`time_zone`::
Defines what time_zone the rollup documents are stored as. Unlike raw data, which can shift timezones on the fly, rolled documents have
to be stored with a specific timezone. By default, rollup documents are stored in `UTC`, but this can be changed with the `time_zone`
parameter.
.Calendar vs Fixed time intervals
**********************************
Elasticsearch understands both "calendar" and "fixed" time intervals. Fixed time intervals are fairly easy to understand;
`"60s"` means sixty seconds. But what does `"1M` mean? One month of time depends on which month we are talking about,
some months are longer or shorter than others. This is an example of "calendar" time, and the duration of that unit
depends on context. Calendar units are also affected by leap-seconds, leap-years, etc.
This is important because the buckets generated by Rollup will be in either calendar or fixed intervals, and will limit
how you can query them later (see <<rollup-search-limitations-intervals, Requests must be multiples of the config>>.
We recommend sticking with "fixed" time intervals, since they are easier to understand and are more flexible at query
time. It will introduce some drift in your data during leap-events, and you will have to think about months in a fixed
quantity (30 days) instead of the actual calendar length... but it is often easier than dealing with calendar units
at query time.
Multiples of units are always "fixed" (e.g. `"2h"` is always the fixed quantity `7200` seconds. Single units can be
fixed or calendar depending on the unit:
[options="header"]
|=======
|Unit |Calendar |Fixed
|millisecond |NA |`1ms`, `10ms`, etc
|second |NA |`1s`, `10s`, etc
|minute |`1m` |`2m`, `10m`, etc
|hour |`1h` |`2h`, `10h`, etc
|day |`1d` |`2d`, `10d`, etc
|week |`1w` |NA
|month |`1M` |NA
|quarter |`1q` |NA
|year |`1y` |NA
|=======
For some units where there are both fixed and calendar, you may need to express the quantity in terms of the next
smaller unit. For example, if you want a fixed day (not a calendar day), you should specify `24h` instead of `1d`.
Similarly, if you want fixed hours, specify `60m` instead of `1h`. This is because the single quantity entails
calendar time, and limits you to querying by calendar time in the future.
**********************************
===== Terms
The `terms` group can be used on `keyword` or numeric fields, to allow bucketing via the `terms` aggregation at a later point. The `terms`
group is optional. If defined, the indexer will enumerate and store _all_ values of a field for each time-period. This can be potentially
costly for high-cardinality groups such as IP addresses, especially if the time-bucket is particularly sparse.
While it is unlikely that a rollup will ever be larger in size than the raw data, defining `terms` groups on multiple high-cardinality fields
can effectively reduce the compression of a rollup to a large extent. You should be judicious which high-cardinality fields are included
for that reason.
The `terms` group has a single parameter:
`fields` (required)::
The set of fields that you wish to collect terms for. This array can contain fields that are both `keyword` and numerics. Order
does not matter
===== Histogram
The `histogram` group aggregates one or more numeric fields into numeric histogram intervals. This group is optional
The `histogram` group has a two parameters:
`fields` (required)::
The set of fields that you wish to build histograms for. All fields specified must be some kind of numeric. Order does not matter
`interval` (required)::
The interval of histogram buckets to be generated when rolling up. E.g. `5` will create buckets that are five units wide
(`0-5`, `5-10`, etc). Note that only one interval can be specified in the `histogram` group, meaning that all fields being grouped via
the histogram must share the same interval.
[[rollup-metrics-config]]
==== Metrics Config
After defining which groups should be generated for the data, you next configure which metrics should be collected. By default, only
the doc_counts are collected for each group. To make rollup useful, you will often add metrics like averages, mins, maxes, etc.
Metrics are defined on a per-field basis, and for each field you configure which metric should be collected. For example:
[source,js]
--------------------------------------------------
"metrics": [
{
"field": "temperature",
"metrics": ["min", "max", "sum"]
},
{
"field": "voltage",
"metrics": ["avg"]
}
]
--------------------------------------------------
// NOTCONSOLE
This configuration defines metrics over two fields, `"temperature` and `"voltage"`. For the `"temperature"` field, we are collecting
the min, max and sum of the temperature. For `"voltage"`, we are collecting the average. These metrics are collected in a way that makes
them compatible with any combination of defined groups.
The `metrics` configuration accepts an array of objects, where each object has two parameters:
`field` (required)::
The field to collect metrics for. This must be a numeric of some kind
`metrics` (required)::
An array of metrics to collect for the field. At least one metric must be configured. Acceptable metrics are min/max/sum/avg/value_count.
.Averages aren't composable?!
**********************************
If you've worked with rollups before, you may be cautious around averages. If an average is saved for a 10 minute
interval, it usually isn't useful for larger intervals. You cannot average six 10-minute averages to find a
hourly average (average of averages is not equal to the total average).
For this reason, other systems tend to either omit the ability to average, or store the average at multiple intervals
to support more flexible querying.
Instead, the Rollup feature saves the `count` and `sum` for the defined time interval. This allows us to reconstruct
the average at any interval greater-than or equal to the defined interval. This gives maximum flexibility for
minimal storage costs... and you don't have to worry about average accuracies (no average of averages here!)
**********************************