2.8 KiB
Fetching information from multiple secrets
In some use cases, it might be impractical to bundle all sensitive information into a single secret, or even it is not possible to fully know a given secret name. In such cases, it is possible that a user might need to sync multiple secrets from an external provider into a single Kubernetes Secret. This is possible to be done in external-secrets with the dataFrom.find
option.
!!! note
The secret's contents as defined in the provider are going to be stored in the kubernetes secret as a single key. Currently, it's possible to apply a decoding Strategy during a find operation, but only at the secret level (e.g. if a secret is a JSON with some B64 encoded data within, decodingStrategy: Auto
would not decode it)
Fetching secrets matching a given name pattern
To fetch multiple secrets matching a name pattern from a common SecretStore you can apply the following manifest:
{% include 'getallsecrets-find-by-name.yaml' %}
Suppose we contain secrets /path/key1
, key2/path
, and path/to/keyring
with their respective values. The above YAML will produce the following kubernetes Secret:
_path_key1: Cg==
key2_path: Cg==
path_to_keyring: Cg==
Fetching secrets matching a set of metadata tags
To fetch multiple secrets matching a name pattern from a common SecretStore you can apply the following manifest:
{% include 'getallsecrets-find-by-tags.yaml' %}
This will match any secrets containing all of the metadata labels in the tags
parameter. At least one tag must be provided in order to allow finding secrets by metadata tags.
Searching only in a given path
Some providers support filtering out a find operation only to a given path, instead of the root path. In order to use this feature, you can pass find.path
to filter out these secrets into only this path, instead of the root path.
Avoiding name conflicts
By default, kubernetes Secrets accepts only a given range of characters. Find
operations will automatically replace any not allowed character with a _
. So if we have a given secret a_c
and a/c
would lead to a naming conflict.
If you happen to have a case where a conflict is happening, you can use the rewrite
block to apply a regexp on one of the find operations (for more information please refer to Rewriting Keys from DataFrom).
You can also set dataFrom.find.conversionStrategy: Unicode
to reduce the collistion probability. When using Unicode
, any invalid character will be replaced by its unicode, in the form of _UXXXX_
. In this case, the available kubernetes keys would be a_c
and a_U2215_c
, hence avoiding most of possible conflicts.
!!! note "PRs welcome" Some providers might not have the implementation needed for fetching multiple secrets. If that's your case, please feel free to contribute!