README.md

Guppy

Server that support GraphQL queries on data from elasticsearch.

Please see this doc for syntax Guppy supports.

Run npm start to start server at port 80.

Local Deployment and Development:

Guppy has some helper script to help a developer to set up a local ES service using Docker, generate some example ES indices for testing, and pop mock data into these example ES indices. Please refer to the DEV Helper doc for more information.

Quickstart with Helm

You can now deploy individual services via Helm! Please refer to the Helm quickstart guide HERE (https://github.com/uc-cdis/guppy/blob/master/doc/quickstart_helm.md)

Configurations:

Before launch, we need to write config and tell Guppy which elasticsearch indices and which auth control field to use. You could put following as your config files:

{
  "indices": [
    {
      "index": "${ES_INDEX_1}",
      "type": "${ES_DOC_TYPE_1}",
      "tier_access_level": "${ES_TIER_ACCESS_LEVEL_1}" // optional, set this if there is no global tierAccessLevel
    },
    {
      "index": "${ES_INDEX_2}",
      "type": "${ES_DOC_TYPE_2}",
      "tier_access_level": "${ES_TIER_ACCESS_LEVEL_2}"  // optional, set this if there is no global tierAccessLevel
    },
    ...
  ],
  "config_index": "${ES_ARRAY_CONFIG}", // optional, if there's array field, Guppy read the configs from this index.
  "auth_filter_field": "${AUTH_FILTER_FIELD}",
  "aggs_include_missing_data": true, // optional, by default true, this boolean decide whether elasticsearch aggregation should return missing data in result
  "missing_data_alias": "no data", // optional, only valid if `aggs_include_missing_data` is true, guppy will alias missing data into this keyword during aggregation. By default it's set to `no data`.
}

Note: Guppy expects that either all indices in the guppy config block will have a tier_access_level set OR that a site-wide TIER_ACCESS_LEVEL is set as an environment variable (or in the global block of a commons' manifest). Guppy will throw an error if the config settings do not meet one of these two expectations. See doc/index_scoped_tiered_access.md for more information.

Following script will start server using at port 3000, using config file example_config.json:

export GUPPY_PORT=3000
export GUPPY_CONFIG_FILEPATH=./example_config.json
npm start

Authorization:

Guppy connects Arborist for authorization. The auth_filter_field item in your config file is the field used for authorization. You could set the endpoint by:

export GEN3_ARBORIST_ENDPOINT=${arborist_service}

If not set, it would default to http://arborist-service. You could set it to mock to skip all authorization steps. But if you just want to mock your own authorization behavior for local test without Arborist, just set INTERNAL_LOCAL_TEST=true. Please look into /src/server/auth/utils.js for more details.

Tiered Access:

The tiered-access setting is configured through either the TIER_ACCESS_LEVEL environment variable or the tier_access_level properties on individual indices in the esConfig. Guppy supports 3 different levels of tiered access:

• private by default: only allows access to authorized resources
• regular: allows all kind of aggregation (with limitation for unauthorized resources), but forbid access to raw data without authorization
• libre: access to all data

For the regular level, there's another configuration environment variable TIER_ACCESS_LIMIT, which is the minimum visible count for aggregation results.

regular level commons can also take in a whitelist of values that won't be encrypted. It is set by config.encrypt_whitelist. By default the whitelist contains missing values: ['__missing__', 'unknown', 'not reported', 'no data']. Also the whitelist is disabled by default due to security reasons. If you would like to enable whitelist, simply put enable_encrypt_whitelist: true in your config. For example, a regular leveled commons with config that looks like this will skip encrypting the value do-not-encrypt-me even if its count is less than TIER_ACCESS_LIMIT:

{
  "indices": [
    {
      "index": "gen3-dev-subject",
      "type": "subject"
    },
    {
      "index": "gen3-dev-file",
      "type": "file"
    }
  ],
  "config_index": "gen3-dev-config",
  "auth_filter_field": "gen3_resource_path",
  "enable_encrypt_whitelist": true,
  "encrypt_whitelist": [ "do-not-encrypt-me" ]
}

The following script will start a Guppy server with a site-wide regular tier access level, and minimum visible count set to 100:

export TIER_ACCESS_LEVEL=regular
export TIER_ACCESS_LIMIT=100
npm start

To learn how to configure Guppy's tiered-access system using a per-index scoping, and which use cases might warrant such a configuration, please see doc/index_scoped_tiered_access.md.

Tier Access Sensitive Record Exclusion

It is possible to configure Guppy to hide some records from being returned in _aggregation queries when Tiered Access is enabled (tierAccessLevel: "regular"). The purpose of this is to "hide" information about certain sensitive resources, essentially making this an escape hatch from Tiered Access. Crucially, Sensitive Record Exclusion only applies to records which the user does not have access to. If the user has access to a record, it will be counted in the aggregation query whether or not it is sensitive.

To enable Sensitive Record Exclusion, set guppy.tier_access_sensitive_record_exclusion_field: "fieldname" in the commons' manifest.json. "fieldName" should match a boolean field in the Elasticsearch index that indicates whether or not a record is sensitive.

(E.g., "tier_access_sensitive_record_exclusion_field": "sensitive" in the Guppy config tells Guppy to look for a field in the ES index called sensitive, and to exclude records in the ES index which have sensitive: "true")

Additional Guppy Endpoints:

Guppy has a special endpoint /download for just fetching raw data from elasticsearch. This endpoint can be used to overcome Elastic Search's 10k record limit. Please see here for details.

Guppy's /_status endpoint yields health check and array field information. This endpoint is publicly accessible and returns output of the form

{"statusCode":200,"warnings":null,"indices":{"<index-name>":{"aliases":{"alias-name":{}},"arrayFields":["<name-of-array-field>"]}}}

The /_version endpoint yields version and commit information. This endpoint is publicly accessible and returns output of the form

{"version":"<version-string>","commit":"<commit-hash>"}