Changelog for Search Guard 7.x-35.0.0-rc1

Release Date: 30.04.2019

Changes

BREAKING: Support for document types and tribe nodes removed

Elasticsearch 7 does no longer support document types and tribe nodes. We removed those features now also from Search Guard.

BREAKING: Deprecated endpoints from REST management API removed

The following endpoints were removed:

/_searchguard/api/actiongroup/
# Use /_searchguard/api/actiongroups/ instead

/_searchguard/api/configuration/{configname}
# Use /_searchguard/api/{configname}/ instead

/_searchguard/api/user/
# Use /_searchguard/api/internalusers/ instead

BREAKING: New configuration format

With Search Guard 7 we introduce a new configuration format to fix a few longstanding issues and to remove the document type level security from it.

The new format is documented here:

Notable changes and improvements at a glance:

  • The new format does now allow dots (.) in index names and regular expressions
  • Every file now has mandatory _sg_meta header
  • Structure and wording is now more precise, aligned and extensible
  • Additional config file for tenants (sg_tenants.yml)

If you do a migration from Elasticsearch/Search Guard 6.x please refer to the upgrade instructions to learn how to migrate the configuration to it’s new format.

If you are using the REST API, you will likely need to change the payload processing of your calls slightly.

For the final release we will also provide a standalone tool to migrate the configuration offline (migrate.sh).

BREAKING: Config format validation

Search Guard 6 and before the configuration syntax check were lenient. This can easily lead to misconfiguration. With Search Guard 7 there is now a strict config syntax validation which rejects invalid syntax as well as configuration parameters which do not exist. The REST API and sgadmin will automatically apply these checks.

But in case you are upgrading from 6.x you have to do this one time manually. Please refer to the upgrade instructions in this case. If you miss this step your cluster can become uninitialized which will result in a downtime.

BREAKING: Default changed for snapshot/restore handling

The default of searchguard.enable_snapshot_restore_privilege changed from false to true (in elasticsearch.yml). This will allow restore operations also for regular users without an admin certificate as long as the snapshot does not contain the global cluster state or the searchguard index. See Enabling snapshot and restore for regular users for more details.

BREAKING: Default changed for permissions evaluation across different roles (multi-rolespan)

Before 7.x-35.0.0-rc1 the default was to grant permissions only, if a single role contained all grants necessary to allow a request. Since 6.x-22.3 there is the multi_rolespan_enabled config option in sg_config.yml to override this behaviour. The default now changes from multi_rolespan_enabled: false to multi_rolespan_enabled: true. This means that a request will be permitted when the combination of all grants from all roles, the user is assigned to, are sufficient to allow access.

BREAKING: Default changed for XFF (x-forwarded-for) support

The default changed from xff.enabled: true to xff.enabled: false (in sg_config.yml)

New feature: Built-in configuration resources

Search Guard 7 ships with built-in roles and actiongroups. Those are static and unchangeable and will be maintained from release to release by us. We strongly recommend to use them where ever possible and prefer them over your own roles and actiongroups. Albeit recommended their usage is not mandatory.

All built-in resource are uppercase and start with SGS_.

New feature: Tenant endpoint for REST management API

For the new tenant file there is now also a corresponding tenant REST API endpoint /_searchguard/api/tenants/. Please refer to the documentation for any details.

New feature: New options for sgadmin to support migration and backups

sgadmin now has this additional command line options:

Known Issues

There is a known issue in this release with regards to Kibana multitenancy when upgrading from Elasticsearch 6. After upgrading the cluster the first attempt to start Kibana 7 may fail with an “index template missing exception” error message. In this case, the workaround is to stop Kibana and start it again. The second attempt to start Kibana 7 will then be successful.