An OVSDB Library written in Go
OVSDB is the Open vSwitch Database Protocol. It's defined in RFC 7047 It's used mainly for managing the configuration of Open vSwitch and OVN, but it could also be used to manage your stamp collection. Philatelists Rejoice!
The API to interact with OVSDB is based on tagged golang structs. We call it a Model. e.g:
type MyLogicalSwitch struct {
UUID string `ovsdb:"_uuid"` // _uuid tag is mandatory
Name string `ovsdb:"name"`
Ports []string `ovsdb:"ports"`
Config map[string]string `ovsdb:"other_config"`
}
libovsdb is able to translate a Model in to OVSDB format. To make the API use go idioms, the following mappings occur:
- OVSDB Set with min 0 and max unlimited = Slice
- OVSDB Set with min 0 and max 1 = Pointer to scalar type
- OVSDB Set with min 0 and max N = Array of N
- OVSDB Enum = Type-aliased Enum Type
- OVSDB Map = Map
- OVSDB Scalar Type = Equivalent scalar Go type
A Open vSwitch Database is modeled using a ClientDBModel which is a created by assigning table names to pointers to these structs:
dbModelReq, _ := model.NewClientDBModel("OVN_Northbound", map[string]model.Model{
"Logical_Switch": &MyLogicalSwitch{},
})
Finally, a client object can be created:
ovs, _ := client.Connect(context.Background(), dbModelReq, client.WithEndpoint("tcp:172.18.0.4:6641"))
client.MonitorAll(nil) // Only needed if you want to use the built-in cache
Once the client object is created, a generic API can be used to interact with the Database. Some API calls can be performed on the generic API: List
, Get
, Create
.
Others, have to be called on a ConditionalAPI
(Update
, Delete
, Mutate
). There are three ways to create a ConditionalAPI
:
Where(): Where()
can be used to create a ConditionalAPI
based on the index information that the provided Models contain. Example:
ls := &LogicalSwitch{UUID: "foo"}
ls2 := &LogicalSwitch{UUID: "foo2"}
ops, _ := ovs.Where(ls, ls2).Delete()
It will check the field corresponding to the _uuid
column as well as all the other schema-defined or client-defined indexes in that order of priority.
The first available index will be used to generate a condition.
WhereAny(): WhereAny()
can be used to create a ConditionalAPI
using a list of Condition objects. Each condition object specifies a field using a pointer
to a Model's field, a ovsdb.ConditionFunction
and a value. The type of the value depends on the type of the field being mutated. Example:
ls := &LogicalSwitch{}
ops, _ := ovs.WhereAny(ls, client.Condition{
Field: &ls.Config,
Function: ovsdb.ConditionIncludes,
Value: map[string]string{"foo": "bar"},
}).Delete()
The resulting ConditionalAPI
will create one operation per condition, so all the rows that match any of the specified conditions will be affected.
WhereAll(): WhereAll()
behaves like WhereAny()
but with AND semantics. The resulting ConditionalAPI
will put all the
conditions into a single operation. Therefore the operation will affect the rows that satisfy all the conditions.
WhereCache(): WhereCache()
uses a function callback to filter on the local cache. It's primary use is to perform cache operations such as
List()
. However, it can also be used to create server-side operations (such as Delete()
, Update()
or Delete()
). If used this way, it will
create an equality condition (using ovsdb.ConditionEqual
) on the _uuid
field for every matching row. Example:
lsList := []LogicalSwitch{}
ovs.WhereCache(
func(ls *MyLogicalSwitch) bool {
return strings.HasPrefix(ls.Name, "ext_")
}).List(&lsList)
The table is inferred from the type that the function accepts as only argument.
The client will track schema indexes and use them when appropriate in Get
, Where
, and WhereAll
as explained above.
Additional indexes can be specified for a client instance to track. Just as schema indexes, client indexes are specified in sets per table. where each set consists of the columns that compose the index. Unlike schema indexes, a key within a column can be addressed if the column type is a map.
Client indexes are leveraged through Where
, and WhereAll
. Since client indexes value uniqueness is not enforced as it happens with schema indexes,
conditions based on them can match multiple rows.
Indexed based operations generally provide better performance than operations based on explicit conditions.
As an example, where you would have:
// slow predicate run on all the LB table rows...
ovn.WhereCache(func (lb *LoadBalancer) bool {
return lb.ExternalIds["myIdKey"] == "myIdValue"
}).List(ctx, &results)
can now be improved with:
dbModel, err := nbdb.FullDatabaseModel()
dbModel.SetIndexes(map[string][]model.ClientIndex{
"Load_Balancer": {{Columns: []model.ColumnKey{{Column: "external_ids", Key: "myIdKey"}}}},
})
// connect ....
lb := &LoadBalancer{
ExternalIds: map[string]string{"myIdKey": "myIdValue"},
}
// quick indexed result
ovn.Where(lb).List(ctx, &results)
This package is divided into several sub-packages. Documentation for each sub-package is available at pkg.go.dev:
- client: ovsdb client and API
- mapper: mapping from tagged structs to ovsdb types
- model: model and database model used for mapping
- ovsdb: low level OVS types
- cache: model-based cache
- modelgen: common code-generator functions
- server: ovsdb test server
- database: database related types, interfaces and implementations
- updates: common code to handle model updates
List the content of the database:
var lsList *[]MyLogicalSwitch
ovs.List(lsList)
for _, ls := range lsList {
fmt.Printf("%+v\n", ls)
}
Search the cache for elements that match a certain predicate:
var lsList *[]MyLogicalSwitch
ovs.WhereCache(
func(ls *MyLogicalSwitch) bool {
return strings.HasPrefix(ls.Name, "ext_")
}).List(&lsList)
for _, ls := range lsList {
fmt.Printf("%+v\n", ls)
}
Create a new element
ops, _ := ovs.Create(&MyLogicalSwitch{
Name: "foo",
})
ovs.Transact(ops...)
Get an element:
ls := &MyLogicalSwitch{Name: "foo"} // "name" is in the table index list
ovs.Get(ls)
And update it:
ls.Config["foo"] = "bar"
ops, _ := ovs.Where(ls).Update(&ls)
ovs.Transact(ops...)
Or mutate an it:
ops, _ := ovs.Where(ls).Mutate(ls, ovs.Mutation {
Field: &ls.Config,
Mutator: ovsdb.MutateOperationInsert,
Value: map[string]string{"foo": "bar"},
})
ovs.Transact(ops...)
Update, Mutate and Delete operations need a condition to be specified. Conditions can be created based on a Model's data:
ls := &LogicalSwitch{UUID:"myUUID"}
ops, _ := ovs.Where(ls).Delete()
ovs.Transact(ops...)
They can also be created based on a list of Conditions:
ops, _ := ovs.Where(ls, client.Condition{
Field: &ls.Config,
Function: ovsdb.ConditionIncludes,
Value: map[string]string{"foo": "bar"},
}).Delete()
ovs.Transact(ops...)
ops, _ := ovs.WhereAll(ls,
client.Condition{
Field: &ls.Config,
Function: ovsdb.ConditionIncludes,
Value: map[string]string{"foo": "bar"},
}, client.Condition{
Field: &ls.Config,
Function: ovsdb.ConditionIncludes,
Value: map[string]string{"bar": "baz"},
}).Delete()
ovs.Transact(ops...)
You can also register a notification handler to get notified every time an element is added, deleted or updated from the database.
handler := &cache.EventHandlerFuncs{
AddFunc: func(table string, model model.Model) {
if table == "Logical_Switch" {
fmt.Printf("A new switch named %s was added!!\n!", model.(*MyLogicalSwitch).Name)
}
},
}
ovs.Cache.AddEventHandler(handler)
In this repository there is also a code-generator capable of generating all the Model types for a given ovsdb schema (json) file.
It can be used as follows:
go install github.com/ovn-org/libovsdb/cmd/modelgen
$GOPATH/bin/modelgen -p ${PACKAGE_NAME} -o {OUT_DIR} ${OVSDB_SCHEMA}
Usage of modelgen:
modelgen [flags] OVS_SCHEMA
Flags:
-d Dry run
-o string
Directory where the generated files shall be stored (default ".")
-p string
Package name (default "ovsmodel")
The result will be the definition of a Model per table defined in the ovsdb schema file.
Additionally, a function called FullDatabaseModel()
that returns the ClientDBModel
is created for convenience.
Example:
Download the schema:
ovsdb-client get-schema "tcp:localhost:6641" > mypackage/ovs-nb.ovsschema
Run go generate
cat <<EOF > mypackage/gen.go
package mypackage
// go:generate modelgen -p mypackage -o . ovs-nb.ovsschema
EOF
go generate ./...
In your application, load the ClientDBModel, connect to the server and start interacting with the database:
import (
"fmt"
"github.com/ovn-org/libovsdb/client"
generated "example.com/example/mypackage"
)
func main() {
dbModelReq, _ := generated.FullDatabaseModel()
ovs, _ := client.Connect(context.Background(), dbModelReq, client.WithEndpoint("tcp:localhost:6641"))
ovs.MonitorAll()
// Create a *LogicalRouter, as a pointer to a Model is required by the API
lr := &generated.LogicalRouter{
Name: "myRouter",
}
ovs.Get(lr)
fmt.Printf("My Router has UUID: %s and %d Ports\n", lr.UUID, len(lr.Ports))
}
To run integration tests, you'll need access to docker to run an Open vSwitch container. Mac users can use boot2docker
export DOCKER_IP=$(boot2docker ip)
docker-compose run test /bin/sh
# make test-local
...
# exit
docker-compose down
By invoking the command make, you will automatically get the same behavior as what is shown above. In other words, it will start the two containers and execute make test-local from the test container.
The libovsdb community is part of ovn-org and can be contacted in the #libovsdb channel in ovn-org Slack server