Retrieve Pillar data by doing a SQL query

This module is not meant to be used directly as an ext_pillar. It is a place to put code common to PEP 249 compliant SQL database adapters. It exposes a python ABC that can be subclassed for new database providers.





Theory of sql_base ext_pillar

Ok, here's the theory for how this works...

  • First, any non-keyword args are processed in order.

  • Then, remaining keywords are processed.

We do this so that it's backward compatible with older configs. Keyword arguments are sorted before being appended, so that they're predictable, but they will always be applied last so overall it's moot.

For each of those items we process, it depends on the object type:

  • Strings are executed as is and the pillar depth is determined by the number of fields returned.

  • A list has the first entry used as the query, the second as the pillar depth.

  • A mapping uses the keys "query" and "depth" as the tuple

You can retrieve as many fields as you like, how they get used depends on the exact settings.

Configuring a sql_base ext_pillar

The sql_base ext_pillar cannot be used directly, but shares query configuration with its implementations. These examples use a fake 'sql_base' adapter, which should be replaced with the name of the adapter you are using.

A list of queries can be passed in

  - sql_base:
      - "SELECT pillar,value FROM pillars WHERE minion_id = %s"
      - "SELECT pillar,value FROM more_pillars WHERE minion_id = %s"

Or you can pass in a mapping

  - sql_base:
      main: "SELECT pillar,value FROM pillars WHERE minion_id = %s"
      extras: "SELECT pillar,value FROM more_pillars WHERE minion_id = %s"

The query can be provided as a string as we have just shown, but they can be provided as lists

  - sql_base:
      - "SELECT pillar,value FROM pillars WHERE minion_id = %s"

Or as a mapping

  - sql_base:
      - query: "SELECT pillar,value FROM pillars WHERE minion_id = %s"
        depth: 2

The depth defines how the dicts are constructed. Essentially if you query for fields a,b,c,d for each row you'll get:

  • With depth 1: {a: {"b": b, "c": c, "d": d}}

  • With depth 2: {a: {b: {"c": c, "d": d}}}

  • With depth 3: {a: {b: {c: d}}}

Depth greater than 3 wouldn't be different from 3 itself. Depth of 0 translates to the largest depth needed, so 3 in this case. (max depth == key count - 1)

Then they are merged in a similar way to plain pillar data, in the order returned by the SQL database.

Thus subsequent results overwrite previous ones when they collide.

The ignore_null option can be used to change the overwrite behavior so that only non-NULL values in subsequent results will overwrite. This can be used to selectively overwrite default values.

  - sql_base:
      - query: "SELECT pillar,value FROM pillars WHERE minion_id = 'default' and minion_id != %s"
        depth: 2
      - query: "SELECT pillar,value FROM pillars WHERE minion_id = %s"
        depth: 2
        ignore_null: True

If you specify as_list: True in the mapping expression it will convert collisions to lists.

If you specify with_lists: '...' in the mapping expression it will convert the specified depths to list. The string provided is a sequence numbers that are comma separated. The string '1,3' will result in:

a,b,c,d,e,1  # field 1 same, field 3 differs
a,b,c,f,g,2  # ^^^^
a,z,h,y,j,3  # field 1 same, field 3 same
a,z,h,y,k,4  # ^^^^
  ^   ^

These columns define list grouping

{a: [
      {c: [
          {e: 1},
          {g: 2}
      {h: [
          {j: 3, k: 4 }

The range for with_lists is 1 to number_of_fields, inclusive. Numbers outside this range are ignored.

If you specify as_json: True in the mapping expression and query only for single value, returned data are considered in JSON format and will be merged directly.

  - sql_base:
      - query: "SELECT json_pillar FROM pillars WHERE minion_id = %s"
        as_json: True

The processed JSON entries are recursively merged in a single dictionary. Additionnaly if as_list is set to True the lists will be merged in case of collision.

For instance the following rows:

{"a": {"b": [1, 2]}, "c": 3} {"a": {"b": [1, 3]}, "d": 4}

will result in the following pillar with as_list=False

{"a": {"b": [1, 3], "c": 3, "d": 4}

and in with as_list=True

{"a": {"b": [1, 2, 3], "c": 3, "d": 4}

Finally, if you pass the queries in via a mapping, the key will be the first level name where as passing them in as a list will place them in the root. This isolates the query results into their own subtrees. This may be a help or hindrance to your aims and can be used as such.

You can basically use any SELECT query that gets you the information, you could even do joins or subqueries in case your minion_id is stored elsewhere. It is capable of handling single rows or multiple rows per minion.

Configuration of the connection depends on the adapter in use.

New in version 3005: The as_json parameter.

More complete example for MySQL (to also show configuration)

  user: 'salt'
  pass: 'super_secret_password'
  db: 'salt_db'

  - mysql:
        query: 'SELECT col1,col2,col3,col4,col5,col6,col7
                  FROM some_random_table
                 WHERE minion_pattern LIKE %s'
        depth: 5
        as_list: True
        with_lists: [1,3]
class salt.pillar.sql_base.SqlBaseExtPillar

This class receives and processes the database rows in a database agnostic way.

as_json = False
as_list = False
depth = 0

Set self.focus for kwarg queries

extract_queries(args, kwargs)

This function normalizes the config block into a set of queries we can use. The return is a list of consistently laid out dicts.

fetch(minion_id, pillar, *args, **kwargs)

Execute queries, merge and return as a dict.

field_names = None
focus = None
ignore_null = False
num_fields = 0
process_fields(field_names, depth)

The primary purpose of this function is to store the sql field list and the depth to which we process.


This function takes a list of database results and iterates over, merging them into a dict form.

result = None
with_lists = None