Elastic Query Filters Challenge - elasticsearch

Question

I have the following query, generating a Top 100 sellers for a given supplier ID, running against a sales index that looks up the product skus for the given supplier in an index of product_skus. This works well.
query = {
size: 0,
query: {
bool: {
filter: [
{
constant_score: {
filter: {
terms: {
sku: {
index: "product_skus",
type: "product",
id: supplier_id,
path: "skus"
}
}
}
}
}
],
must_not: []
}
},
aggs: {
unit_sum: {
terms: {
field: "sku",
size: 100,
order: {
one: "desc"
}
},
aggs: {
one: {
sum: {
field: "units"
}
}
}
}
}
}
Now I have a scenario where a given user needs to have their access restricted to a subset of the suppliers skus. I am trying to get my head around the best way to tackle this. I am leaning towards having another index of the Skus a user can access and doing a second lookup, but I can't quite get my head around the query logic.
In simple terms for example; if in the above query, for supplier 1 we return products [A,B,C,D,E]
and user John should only see the results based on products [A,C,E]
How would I go about writing the query to do this? Is it as simple as adding in a should clause after the filter inside the bool?
Thanks in advance!

Answer 1

Routing is probably what you need in this situation, as your scenario allows you to use routing for users. As an additional bonus of organizing your data into separate shards, it will allow to increase performances when routing is used in query. Why? Because when using routing, the request will be sent only to shards containing the relevant data instead of every node across the cluster.
What would it look like in your case? Let's have a look with a simple mapping, and a product that should only be accessed with an id 123:
The mapping of product_skus (modify as needed):
PUT product_skus
{
"settings": {
"index": {
"number_of_shards": "5",
"number_of_replicas": "1"
}
},
"mappings": {
"product": {
"_routing": {
"required": true
},
"properties": {
"supplierId":{
"type": "integer"
}, "path":{
"type": "string"
}
}
}
}
}
Now let's put a product in the index type (notice the routing):
POST product_skus/product?routing=123
{
"supplierId": 123,
"path": "some/path"
}
And finally two requests and their output using the routing:
GET product_skus/_search?routing=123
{
"query": {
"match_all": {}
}
}
Output:
{
"took": 4,
"timed_out": false,
"_shards": {
"total": 1,
"successful": 1,
"failed": 0
},
"hits": {
"total": 1,
"max_score": 1,
"hits": [
{
"_index": "product_skus",
"_type": "product",
"_id": "AVrMHzgx28yun46LEMYm",
"_score": 1,
"_routing": "123",
"_source": {
"supplierId": 123,
"path": "some/path"
}
}
]
}
}
Second query:
GET product_skus/_search?routing=124
{
"query": {
"match_all": {}
}
}
Output:
{
"took": 1,
"timed_out": false,
"_shards": {
"total": 1,
"successful": 1,
"failed": 0
},
"hits": {
"total": 0,
"max_score": null,
"hits": []
}
}
It is only a simple example, you might want to check the documentation for more information:
The _routing field
Another routing example
An example of routing with fields of the type
In addition the following shows that only one shard is used with routing:
GET product_skus/_search_shards?routing=123
Output:
{
"nodes": {
"1sMKtN6aQ9yyOsTjknWyQA": {
"name": "1sMKtN6",
"ephemeral_id": "X-V2QGTwTmqUFQb1B6KIUw",
"transport_address": "127.0.0.1:9300",
"attributes": {}
}
},
"shards": [
[
{
"state": "STARTED",
"primary": true,
"node": "1sMKtN6aQ9yyOsTjknWyQA",
"relocating_node": null,
"shard": 0,
"index": "product_skus",
"allocation_id": {
"id": "1MMkFaALRxm1N-x8J8AGhg"
}
}
]
]
}
See the search shards API for more details.