Sysrepo is a YANG-based configuration and operational state data store for Unix/Linux applications.
Applications can use sysrepo to store their configuration modeled by provided YANG model instead of using e.g. flat configuration files. Sysrepo will ensure data consistency of the data stored in the datastore and enforce data constraints defined by YANG model. Applications can currently use C language API of sysrepo Client Library to access the configuration in the datastore, but the support for other programming languages is planed for later, too.
Sysrepo can be easily integrated with management agents such as NETCONF or RESTCONF servers, using the same client library API that applications use to access their configuration. As of now, sysrepo is integrated with the Netopeer 2 NETCONF server. This means that applications that use sysrepo to store their configuration can automatically benefit from the ability to being controlled via NETCONF.
The project uses 2 main branches master
and devel
. Other branches should not be cloned. In master
there are files of the
last official release. Any latest improvements and changes, which were tested at least briefly are found in devel
. On every
new release, devel
is merged into master
.
This means that when only stable official releases are to be used, either master
can be used or specific releases downloaded.
If all the latest bugfixes should be applied, devel
branch is the one to be used. Note that whenever a new issue is created
and it occurs on the master
branch, the first response will likely be to use devel
before any further provided support.
When upgrading Sysrepo to a newer major SO version, look into the compatibility
directory for a summary of changes.
Each directory describes the changes between the specific SO versions. Be aware that they do not match project versions.
- Ability to store / retrieve YANG-modeled data elements adressed by XPath
- Startup, running, candidate, and operational datastore support
- Data consistency and constraints enforcement according to YANG models
- No single point of failure design (is just a library)
- Full transaction and concurrency support
- Notifications of subscribed applications about the changes made in the datastore
- Commit verifiers (change verification by subscribed applications)
- Operational data support (publishing of application's state/configuration data to sysrepo)
- YANG 1.1 support
- Custom RPC, Event Notifications, YANG 1.1 Actions support
- Notification store & notification replay
- Factory Default Datastore support (RFC 8808)
Binary RPM or DEB packages of the latest release can be built locally using apkg
, look into README
in
the distro
directory.
Sysrepo does not have any master process that could enforce complex access control. So instead, it relies on and utilizes standard file system permissions but there are some things to bear in mind.
To prevent any sensitive data from being accessible by unauthorized processes, it is imperative to always
set correct permissions and owner for all YANG modules being installed. The utility sysrepoctl
can help
with both displaying all the permissions (--list
) and modifying them (--change <module>
) in addition
to this functionality being available in the API.
Having made certain of this, the default configuration should be suitable for a reasonably secure machine that has no malicious running processes. Specifically, it is trivial for such a process to completely break sysrepo by writing into shared files that must be accessible for all the processes linked with sysrepo. Also, with some reverse engineering, it may even be possible to access data by an unathorized process when they are being communicated in these shared files.
In order to avoid all such security issues, there are 2 cmake
variables SYSREPO_UMASK
and SYSREPO_GROUP
that should be adjusted. Generally, a new system group should be created and set for SYSREPO_GROUP
and then
all outside access frobidden by setting SYSREPO_UMASK
to 00007
. If then all the users executing sysrepo
processes belong to this group, none of sysrepo files and no sensitive information should be accessible to
other users.
- C compiler (gcc >= 4.8.4, clang >= 3.0, ...)
- cmake >= 2.8.12
- libyang
- tar(1)
- pkg-config & libsystemd (to support
sysrepo-plugind
systemd service) - mongodb-org; libmongoc >= 1.24.0; libbson >= 1.24.0 (for MONGO DS datastore plugin)
- redis-stack-server; hiredis >= 1.1.0 (for REDIS DS datastore plugin)
- doxygen (for generating documentation)
- cmocka >= 1.0.1 (for tests only, see Tests)
- valgrind (for enhanced testing)
- gcov; lcov; genhtml (for code coverage)
$ mkdir build; cd build
$ cmake ..
$ make
# make install
Another markdown document aimed at plugin developers is available in plugin_dev_doc.md. The goal of the document is to provide a single place where a complete overview of information required to start developing plugins is available. The document describes the basics of Sysrepo plugin development and the technologies required to work with Sysrepo like YANG, XPath, NETCONF and others.
Set custom repository path:
-DREPO_PATH=/opt/sysrepo/my_repository
Set custom sysrepo
DS and NTF plugins path:
-DSR_PLUGINS_PATH=/opt/sysrepo/plugins
Set custom sysrepo-plugind
plugins path:
-DSRPD_PLUGINS_PATH=/opt/sysrepo-plugind/plugins
Set global umask
for all sysrepo file and directory creation:
-DSYSREPO_UMASK=00007
Set system group to own all sysrepo-related files:
-DSYSREPO_GROUP=sysrepo
Set systemd
system service unit path:
-DSYSTEMD_UNIT_DIR=/usr/lib/systemd/system
Set NACM recovery username with unrestricted access:
-DNACM_RECOVERY_USER=root
Set NACM configuration data and 'sysrepo-monitoring' default permissions:
-DNACM_SRMON_DATA_PERM=000
Set startup
and factory-default
datastore data for internal modules (such as ietf-netconf-acm
):
-DINTERNAL_MODULE_DATA_PATH=/etc/config/factory_default_config.xml
Set CC
variable:
$ CC=/usr/bin/clang cmake ..
To change the prefix where the library, headers and any other files are installed,
set CMAKE_INSTALL_PREFIX
variable:
$ cmake -DCMAKE_INSTALL_PREFIX:PATH=/usr ..
Default prefix is /usr/local
.
There are two build modes:
- Release. This generates library for the production use without any debug information.
- Debug. This generates library with the debug information and disables optimization of the code.
The Debug
mode is currently used as the default one. to switch to the
Release
mode, enter at the command line:
$ cmake -D CMAKE_BUILD_TYPE:String="Release" ..
Based on the tests run, it is possible to generate code coverage report. But it must be enabled and these commands are needed to generate the report:
$ cmake -DENABLE_COVERAGE=ON ..
$ make
$ make coverage
All basic Sysrepo functions are available via the main header:
#include <sysrepo.h>
To compile your program with libsysrepo, it is necessary to link it using the following linker parameters:
-lsysrepo
Note, that it may be necessary to call ldconfig(8)
after library installation and if the
library was installed into a non-standard path, the path to it must be specified to the
linker. To help with setting all the compiler's options, there is sysrepo.pc
file for
pkg-config(1)
available in the source tree. The file is installed with the library.
There are additional utility functions available as part of Sysrepo but their headers need to be included manually. They are:
#include <sysrepo/error_format.h>
Implements basic creation and retrieval of well-known Sysrepo errors generated by callbacks. Currently, the only supported error format is NETCONF.
#include <sysrepo/netconf_acm.h>
NETCONF Access Control Module with configuration data in ietf-netconf-acm YANG module is implemented
as part of sysrepo. By default, no users other than the recovery user (default root
) will be allowed
to write any data but should be granted read and execute permissions unless the access was modified
by a NACM extension.
#include <sysrepo/subscribed_notifications.h>
This header provides additional functionality to simplify implementing ietf-subscribed-notifications and
ietf-yang-push YANG modules. But these modules need to be installed manually and can be found with all
their imports in modules/subscribed_notifications
.
#include <sysrepo/values.h>
Utility header for working with sr_val_t
Sysrepo value structures. Note that these are generally
considered deprecated and libyang struct lyd_node
should be used instead.
#include <sysrepo/xpath.h>
More complex handling of XPath expressions that has lots of features at the cost of efficiency.
It is possible to change the repository path by setting SYSREPO_REPOSITORY_PATH
variable.
Also, if SYSREPO_SHM_PREFIX
is defined, it is used for all SHM files created. This way
several sysrepo instances can effectively be run simultanously on one machine.
SR_ENV_RUN_TESTS
can be used when building packages which run tests that use sysrepo.
This will enable the package tests to run without having priviliges to chown files to SYSREPO_GROUP
.
There are 2 simple binaries sysrepoctl(1)
and sysrepocfg(1)
included that can execute commands related to
managed YANG modules and stored YANG data, respectively. Full CLI is available only as separate projects such as:
- onm-cli (
C
) - netconf-cli (
C++
) - includessysrepo-cli
The factory-default
datastore contents of a module are automatically populated by the initial data used
when installing the specific module and cannot be changed (unless the module is reinstalled). There is
an internal subscription to the /ietf-factory-default:factory-reset
RPC which performs the copying of
factory-default
data into all the other datastores. This RPC has a priority 10 so applications are
able to subscribe to it with higher or lower priority and perform any other tasks required for a device
to be rest to its factory settings.
In sysrepo there are three internal datastore plugins (JSON DS file
, MONGO DS
and REDIS DS
). The default datastore
plugin is JSON DS file
which stores all the data to JSON files. MONGO DS
and REDIS DS
store data to a database and can be used
as the default datastore plugins for various datastores after setting a few CMake
variables. For every datastore a different default datastore plugin can be set. For example:
cmake -DDEFAULT_STARTUP_DS_PLG="MONGO DS" -DDEFAULT_RUNNING_DS_PLG="MONGO DS" -DDEFAULT_CANDIDATE_DS_PLG="REDIS DS" -DDEFAULT_OPERATIONAL_DS_PLG="JSON DS file" -DDEFAULT_FACTORY_DEFAULT_DS_PLG="JSON DS file" ..
The shared memory prefix set by SYSREPO_SHM_PREFIX
is used by each plugin to isolate data between separate sysrepo "instances".
JSON DS file
includes it in the name of every file it creates, whereas MONGO DS
includes it
in the name of every collection and lastly REDIS DS
includes it in the name of every key as a part of the prefix.
For more information about plugins, see plugin documentation.
To use MONGO DS
datastore plugin, libmongoc and libbson libraries have to be present
on the system. Additionally a running MongoDB server has to be available to the system. By default
sysrepo assumes that the server is available at the loopback address 127.0.0.1
and port 27017
with
no authentication needed. For different IP address and port, set MONGO_HOST
and MONGO_PORT
CMake
variables. For the authentication via username and password, set MONGO_USERNAME
and MONGO_PASSWORD
CMake variables. Please note that for sysrepo to correctly authenticate, an existing user with sufficient rights
and with the configured username and password has to be available
on the server. Also if the user is created on a different database than admin
, provide the correct database name on which the user was created
via the MONGO_AUTHSOURCE
CMake variable. Lastly, for the authentication to work, authentication has to be enabled in the server configuration (see Official MongoDB documentation).
For more information on how the plugin works, please refer to the plugin documentation.
Similarly to MONGO DS
, to use REDIS DS
datastore plugin, libhiredis client library and Redis Stack server have to be available to the system.
The default server address 127.0.0.1
and port 6379
are assumed with
no authentication needed. For different IP address and port, set REDIS_HOST
and REDIS_PORT
CMake variables.
To enable authentication via a username and password, set REDIS_USERNAME
and REDIS_PASSWORD
CMake variables,
create a corresponding user with sufficient rights, and do not forget to enforce the authentication on the server (see official Redis documentation).
For more information on how the plugin works, please refer to the plugin documentation.
See examples directory, which contains an example for basic API functions.
There are no bindings for other languages directly in this project but they are available separately.
There are several tests included and built with cmocka. The tests
can be found in tests
subdirectory and they are designed for checking library
functionality after code changes.
The tests are by default built in the Debug
build mode by running
$ make
In case of the Release
mode, the tests are not built by default (it requires
additional dependency), but they can be enabled via cmake option:
$ cmake -DENABLE_TESTS=ON ..
Note that if the necessary cmocka headers are not present in the system include paths, tests are not available despite the build mode or cmake's options.
Tests can be run by the make's test
target:
$ make test
There is a performance measurement tool included that prints information about the time required to execute common use-cases of working with large YANG instance data.
To enable this test, use an option and to get representative results, enable Release build type:
$ cmake -DCMAKE_BUILD_TYPE=Release -DENABLE_PERF_TESTS=ON ..
and to run the test with seeing its output run:
$ make
$ ctest -V -R sr_perf