-
Notifications
You must be signed in to change notification settings - Fork 1.1k
Configuration
SONiC is managing configuration in a single source of truth - a redisDB instance that we refer as ConfigDB. Applications subscribe to ConfigDB and generate their running configuration correspondingly.
(Before Sep 2017, we were using an xml file named Minigraph to configure SONiC devices. For historical documentation, please refer to Configuration with Minigraph)
In current version of SONiC, ConfigDB is implemented as database 4 of local redis. When system boots, configurations will be loaded from /etc/sonic/config_db.json file into redis. Please note that ConfigDB content won't be written back into /etc/sonic/config_db.json file automatically. In order to do that, a config save command need to be manually executed from CLI. Similarly, config load will trigger a force load of json file into DB. Generally, content in /etc/sonic/config_db.json can be considered as starting config, and content in redisDB running config.
We keep a way to load configuration from minigraph and write into ConfigDB for backward compatibility. To do that, run config load_minigraph.
The design of ConfigDB supports incremental configuration - application could subscribe to changes in ConfigDB and response correspondingly. However, this feature is not implemented by all applications yet. By Sep 2017 now, the only application that supports incremental configuration is BGP (docker-fpm-quagga). For other applications, a manual restart is required after configuration changes in ConfigDB.
ConfigDB uses a table-object schema that is similar with AppDB, and config_db.json is a straight-forward serialization of DB. As an example, the following fragments could be BGP-related configuration in redis and json, correspondingly:
127.0.0.1:6379[4]> keys BGP_NEIGHBOR:*
1) "BGP_NEIGHBOR:10.0.0.31"
2) "BGP_NEIGHBOR:10.0.0.39"
3) "BGP_NEIGHBOR:10.0.0.11"
4) "BGP_NEIGHBOR:10.0.0.7"
...
127.0.0.1:6379[4]> hgetall BGP_NEIGHBOR:10.0.0.3
1) "admin_status"
2) "up"
3) "peer_addr"
4) "10.0.0.2"
5) "asn"
6) "65200"
7) "name"
8) "ARISTA07T2"
"BGP_NEIGHBOR": {
"10.0.0.61": {
"local_addr": "10.0.0.60",
"asn": 64015,
"name": "ARISTA15T0"
},
"10.0.0.49": {
"local_addr": "10.0.0.48",
"asn": 64009,
"name": "ARISTA09T0"
},
...
}
Full sample config_db.json files are availables at here and here.
The DEVICE_METADATA table contains only one object named localhost. In this table the device metadata such as hostname, hwsku, deployment envionment id and deployment type are specified. BGP local AS number is also specified in this table as current only single BGP instance is supported in SONiC.
"DEVICE_METADATA": {
"localhost": {
"bgp_asn": 65100,
"deployment_id": null,
"hostname": "switch1",
"type": "LeafRouter",
"hwsku": "Force10-S6000"
}
},
In this table the physical port configurations are defined. Each object will have port name as its key, and port name alias and port speed as optional attributes.
"PORT": {
"Ethernet0": {
"alias": "fortyGigE0/0",
"speed": "40000"
},
"Ethernet4": {
"alias": "fortyGigE0/4"
},
"Ethernet8": {
"alias": "fortyGigE0/8"
},
...
}
Port channels are defined in PORTCHANNEL table with port channel name as object key and member list as attribute.
"PORTCHANNEL": {
"PortChannel02": {
"members": [
"Ethernet116"
]
},
"PortChannel01": {
"members": [
"Ethernet112"
]
},
},
This table is where VLANs are defined. Vlan name is used as object key, and member list as well as an integer id are defined as attributes.
"VLAN": {
"Vlan1000": {
"members": [
"Ethernet0",
"Ethernet4",
"Ethernet8",
"Ethernet12"
],
"vlanid": "1000"
}
},
VLAN member table has Vlan name together with physical port or port channel name as object key, and tagging mode as attributes.
"VLAN_MEMBER": {
"Vlan1000|PortChannel47": {
"tagging_mode": "untagged"
},
"Vlan1000|Ethernet8": {
"tagging_mode": "untagged"
},
"Vlan2000|PortChannel47": {
"tagging_mode": "tagged"
}
},
IP configuration for data plane are defined in INTERFACE, PORTCHANNEL_INTERFACE, and VLAN_INTERFACE table. The objects in all three tables have the interface (could be physical port, port channel, or vlan) that IP address is attached to as first-level key, and IP prefix as second-level key. IP interface objects don't have any attributes.
"INTERFACE": {
"Ethernet0|10.0.0.0/31": {},
"Ethernet4|10.0.0.2/31": {},
"Ethernet8|10.0.0.4/31": {},
...
},
"PORTCHANNEL_INTERFACE": {
"PortChannel01|10.0.0.56/31": {},
"PortChannel01|FC00::71/126": {},
"PortChannel02|10.0.0.58/31": {},
"PortChannel02|FC00::75/126": {}
...
},
"VLAN_INTERFACE": {
"Vlan1000|192.168.0.1/27": {}
},
Loopback interface configuration lies in LOOPBACK_INTERFACE table and has similar schema with data plane interfaces. The loopback device name and loopback IP prefix act as multi-level key for loopback interface objects.
"LOOPBACK_INTERFACE": {
"Loopback0|10.1.0.32/32": {},
"Loopback0|FC00:1::32/128": {}
},
Management interfaces are defined in MGMT_INTERFACE table. Object key is composed of management interface name and IP prefix. Attribute gwaddr specify the gateway address of the prefix. forced_mgmt_routes attribute can be used to specify addresses / prefixes traffic to which are forced to go through management network instead of data network.
"MGMT_INTERFACE": {
"eth0|10.3.100.3/23": {
"forced_mgmt_routes": [
"10.0.10.0/29",
"10.0.20.5"
],
"gwaddr": "10.3.100.1"
}
},
BGP session configuration is defined in BGP_NEIGHBOR table. BGP neighbor address is used as key of bgp neighbor objects. Object attributes include remote AS number, neighbor router name, and local peering address. Dynamic neighbor is also supported by defining peer group name and IP ranges in BGP_PEER_RANGE table.
"BGP_NEIGHBOR": {
"10.0.0.61": {
"local_addr": "10.0.0.60",
"asn": 64015,
"name": "ARISTA15T0"
},
"10.0.0.49": {
"local_addr": "10.0.0.48",
"asn": 64009,
"name": "ARISTA09T0"
},
...
},
"BGP_PEER_RANGE": {
"PeerGroup": {
"name": "PeerGroup",
"ip_range": [
"192.168.0.0/27"
]
}
},
The L2 neighbor and connection information can be configured in DEVICE_NEIGHBOR table. Those information are used mainly for LLDP. While mandatory fields include neighbor name acting as object key and remote port / local port information in attributes, optional information about neighbor device such as device type, hwsku, management address and loopback address can also be defined.
"DEVICE_NEIGHBOR": {
"ARISTA04T1": {
"mgmt_addr": "10.20.0.163",
"hwsku": "Arista",
"lo_addr": null,
"local_port": "Ethernet124",
"type": "LeafRouter",
"port": "Ethernet1"
},
"ARISTA03T1": {
"mgmt_addr": "10.20.0.162",
"hwsku": "Arista",
"lo_addr": null,
"local_port": "Ethernet120",
"type": "LeafRouter",
"port": "Ethernet1"
},
"ARISTA02T1": {
"mgmt_addr": "10.20.0.161",
"hwsku": "Arista",
"lo_addr": null,
"local_port": "Ethernet116",
"type": "LeafRouter",
"port": "Ethernet1"
},
"ARISTA01T1": {
"mgmt_addr": "10.20.0.160",
"hwsku": "Arista",
"lo_addr": null,
"local_port": "Ethernet112",
"type": "LeafRouter",
"port": "Ethernet1"
}
},
These information are configured in individual tables. Domain name or IP address of the server is used as object key. Currently there is no attributes in those objects.
"NTP_SERVER": {
"2.debian.pool.ntp.org": {},
"1.debian.pool.ntp.org": {},
"3.debian.pool.ntp.org": {},
"0.debian.pool.ntp.org": {}
},
"DHCP_SERVER": {},
"SYSLOG_SERVER": {}，
ACL and mirroring related configuration are defined in MIRROR_SESSION, ACL_TABLE and ACL_RULE tables. Those tables are in progress of migrating from APPDB. Please refer to their schema in APPDB here and migration plan here.
"MIRROR_SESSION": {
"everflow0": {
"src_ip": "10.1.0.32",
"dst_ip": "2.2.2.2"
}
},
"ACL_TABLE": {
"data_acl": {
"policy_desc" : "data_acl",
"type": "l3",
"port": [
"Ethernet0",
"Ethernet4",
"Ethernet8",
"Ethernet12"
],
}
},
To be added.
Detail instruction to be added. A sample could be found in this PR that implemented dynamic configuration for BGP.
-
For Users
-
For Developers
-
Subgroups/Working Groups
-
Presentations
-
Join Us