This PR brings back the built-in tree of downloaded RISwhois data rather
than using the Roto API. It does so using a memory-optimized tree
implementation and has a much smaller memory footprint than the previous
iteration. At the time of writing, the a full RISwhois dataset requires 55
megabytes of memory.
This PR also reverts the changes to the configuration. It removes the
bgp_api_enabled, bgp_api_uri, and bgp_api_cache_duration fields and adds
bgp_riswhois_enabled, bgp_riswhois_v4_uri, bgp_riswhois_v6_uri, and
bgp_riswhois_refresh_duration fields, all of which are optional.
Because of these config changes, the PR is a breaking change.
This changes the feedback provided by krillta proxy signer show-request to
include information about certificate expiration and renewal time.
---------
Co-authored-by: Koen van Hove <koen@nlnetlabs.nl>
This PR adds caching of responses received from the BGP API. The caching
duration can be set via the new config variable `bgp_api_cache_seconds`
which defaults to 1800, i.e., 30 minutes.
Initially, this PR was intended to more clearly separate code between Krill
“business logic” – now bundled in a KrillManager –, and the HTTP server code
that serves the API. The former now lives in the server module, the latter
in the daemon module together with all the code to spin up a Krill daemon
driving the HTTP server.
However, along the way it turned into a complete redesign of how the HTTP
server code works. Request handling has been split into three stages that
forces implementers to check for permissions (or actively choose to not
check), and read the body (or check that there isn’t one). Dispatching of
the request has been restructured which should make it easier to follow what
goes on where.
This PR increases the minimum Rust version to 1.81.
This PR renames the daemon module into the server module. This is in
preparation of separating the “business logic” from the actual daemon
driving it (which will then live in a new daemon module).
This PR refactors and reorganizes a lot of code. It attempts to organize
things in a way that allows to keep more things private. Conversely, it
made the fields of many data-only structs pub and removed the creator,
accessor, and unpack methods for those.
The PR deliberately avoid any functional code changes given that due to
its sheer size, it is essentially unreviewable.
This is also why it stopped short of re-organizing the structure of the
actual server, ie., the Krillserver and the daemon::http module which
currently have a rather blurry distinction. A follow-up PR will create a
more clear separation but this requires code changes.
This PR restructures how authentication policies are used in Krill. It
removes the use of Oso and its policy definition language and instead
switches to simple, straightforward mappings between permissions, roles,
and users.
The existing concept of roles is augmented to serve as the central
configuration option for limiting a user’s access to certain action and
resources. Roles are now user configurable via the new auth_roles
configuration directive. For each role, a set of permissions has to be
provided. Optionally, a list of resource handles (vulgo: CAs) can be given
in which case access is limited to these resources.
The authentication providers now assign one of these roles to each logged
in user.
The OpenID Connect provider now only determines claims for “id,” i.e., the
user name, and the “role.” Since we replaced the previous use of JMES
paths with custom functions with a more stringent model of matching and
substitution, the configuration had to change in a non-compatible way,
anyway, so we cleaned it up a bit and switched from a map to an array for
the claims.
For the config file provider, this was already possible by adding a “role”
attribute. This has now been changed into a “role” field of the user
details. In order to make upgrading seamless, the “role” attribute is
still accepted but a deprecation warning is logged. Since the auth_users
configuration is not used for the OpenID Connect provider any more, the
password_hash and salt fields of the user details are now mandatory.
Custom policies have been removed.
This is a breaking change.
This PR changes how the clients -- krillc, krillta, as well as the
integration tests -- work to better fit the derive model provided by clap.
This results in basically everything in the cli module and all the
integration tests being different now.
The PR slightly changes the options for both krillc and krillta. For krillc,
the --server, --token, --format, and --api options are now before the first
subcommand (since they affect all commands). For krillta, those options are
now after krillta proxy but before the next subcommand, while --format is
now after krillta signer.
This PR also removes client support and integration tests for RTA.
This is a breaking change.
* Add mapping to support that parent and child use different rc names.
* Export child (so that we can test importing it).
* Import child.
* Force parent refresh in resource check loop.
* allow users to just use this, rather than create/delete
* needed to deal with potentially existing history where Krill
instances <0.13.0 ended up with AspaDefinitions with an
empty providers list.
* Remove support for migrating pre-0.9.0 versions.
* Keep publisher files in one place only.
* Improve merging new publish elements into staged elements.
* Upgrade pre-0.12.0 RepositoryContent.
* Upgrade v0.12.x RepositoryContent
* Log errors in case of merge conflicts
* No session reset on 0.13 upgrade (#1028)
---------
Co-authored-by: ximon18 <3304436+ximon18@users.noreply.github.com>
Martin Hoffmann