# Metric Persistence Filter

The metric persistence filter, configurable in the Metric Editor, can be used to discard incoming series commands according to a filter expression. Commands for which the expression returns false are not stored in the database.

# Expression Syntax

The filter is a boolean condition that can include fields, operators, and functions.

# Fields

Name	Type	Description	Example
entity	string	Entity name.	entity LIKE "?tsd"
value	number	Numeric value.	value > 0
message	string	Text value (annotation).	message = "a"
timestamp	number	series command timestamp.	timestamp < 1522683114614
tags.{name} or tags['name']	string	Value of command tag with name name. Tag names are case-insensitive.	tags.location NOT IN ('a', 'b', 'c') tags['fs'] LIKE "ext*".

# Operators

Comparison operators: =, ==, !=, LIKE.

Logical operators: AND, OR, NOT as well as && , ||, !

# Wildcards

• Wildcard * means zero or more characters.
• Wildcard ? means any character.

# Functions

The following functions can be used in the filter expression:

• collection
• list
• likeAll
• likeAny
• matches
• startsWithAny
• contains
• collection_contains
• collection_intersects
• upper
• lower
• size
• isEmpty
• IN

Filter expression can also include Math functions.

Math.sin(Math.toRadians(value)) < 0.5

# collection

collection(string s) [string]

Returns an array of strings which have been loaded with the specified string s.

Named collections are listed on the Data > Named Collections page.

To check the size of the collection, use the .size() method.

To access the n-th element in the collection, use square brackets as in [index] or the get(index) method (starting with 0 for the first element).

entity = collection('hosts')[0]

tags.request_ip = collection('ip_white_list').get(1)

# list

list(string s[, string p]) [string]

Splits string s using separator p (default is comma ',') into a collection of string values. The function discards duplicate items by preserving only the first occurrence of each element.

To access the n-th element in the collection, use square brackets as in [index] or the get(index) method (starting with 0 for the first element).

Examples:

entity = list('atsd,nurswgvml007').get(0)

# likeAll

likeAll(object s, [string] c) boolean

Returns true, if the first argument s matches every element in the collection of patterns c. The collection c can be specified inline as an array of strings or reference a named collection.

Examples:

likeAll(tags.request_ip, ['192.0.*', '192.0.2.?'])

# likeAny

likeAny(object s, [string] c) boolean

Returns true, if the first argument s matches at least one element in the collection of patterns c. The collection c can be specified inline as an array of strings or reference a named collection.

Examples:

likeAny(tags.os, ['Ubuntu*', 'CentOS*'])

likeAny(tags.request_ip, collection('ip_white_list'))

# matches

matches(string p, [string] c) boolean

Returns true if one of the elements in collection c matches (satisfies) the specified pattern p. The collection c can be specified inline as an array of strings or reference a named collection.

The pattern supports wildcard characters asterisk (*) and question mark (?).

Example:

matches(entity, collection('hosts'))

matches(message, ['OK', 'stable'])

# startsWithAny

startsWithAny(object s, [string] c) boolean

Returns true, if the first argument s starts with any of strings from collection c. Collection c can be specified inline as an array of strings or reference a named collection.

Examples:

startsWithAny(entity, ['a', 'nur'])

# contains

[string].contains(string s) boolean

Returns true if s is contained in the collection. The collection c can be specified inline as an array of strings or reference a named collection.

Example:

collection('ip_white_list').contains(tags.request_ip)

# collection_contains

collection_contains(object v, [] c) boolean

Returns true, if collection c contains object v.

Examples:

collection_contains(os_name, ['linux', 'unix'])

An overloaded function that accepts the name and value and returns true if the specified collection contains the value.

collection_contains(string value, string name) boolean

collection_contains(os_name, 'os_whitelist')

The match is case-sensitive in both cases.

# collection_intersects

collection_intersects([] f, [] s) boolean

Returns true, if collection f has elements in common with collection s. The collections can be specified inline as an arrays of strings or reference a named collections.

Examples:

collection_intersects(tags.values(), collection('ip_white_list'))

# upper

upper(string s) string

Converts s to uppercase letters.

# lower

lower(string s) string

Converts s to lowercase letters.

# size

[].size() integer

Returns the number of elements in the collection.

Examples:

tags.size() > 1

# isEmpty

[].isEmpty() boolean

Returns true if the number of elements in the collection is zero.

Example:

NOT tags.isEmpty()

# IN

string s IN (string a[, string b[...]]) boolean

Returns true if s is contained in the collection of strings enclosed in round brackets.

Examples:

entity IN ('nurswgvml007', 'nurswgvml008')

tags.location IN ('NUR', 'SVL')

# Using Persistence Filter

To reduce disk space utilization, open the Settings > Receive Statistics page to view metrics with the highest number of inserted commands.

The Series icon opens a list of unique tags for the selected metric.

If some of the incoming data is of low value it can be ignored when received.

To stop storing such series, create a collection with filter patterns on the Data > Named Collections page.

Open metric editor and enter a filter expression to ignore matching series.

!likeAny(tags.command, collection('ignore-collector-process-commands'))

As a result, the number of stored series is reduced.