Architecture

Overview #

VTAdmin is made up of two components:

  • vtadmin-web, the browser interface
  • vtadmin-api, the HTTP(S) and gRPC API

The vtadmin-web front-end queries its data from vtadmin-api. In turn, vtadmin-api issues queries against the vtgates and vtctlds across one or more Vitess clusters. The clusters.yaml config file defines this mapping from VTAdmin to cluster(s).

Single-cluster deployments are the simplest and most common configuration. The local example in the "Get Started" documentation is a good example of a single-cluster deployment.

flowchart LR vtadmin-web --> vtadmin-api vtadmin-api --> vtgate vtadmin-api --> vtctld vtgate <--> topology vtctld <--> topology topology["Topology Service"]

Large Vitess deployments can be "multi-cluster". These environments have two or more Vitess clusters, each with its own Topology Service, that are isolated from one another. VTAdmin supports these, too. An example of a multi-cluster Vitess deployment is an environment that has one Vitess cluster per geographic region:

flowchart LR vtadmin-web --> vtadmin-api vtadmin-api --> vtgate_us vtadmin-api --> vtctld_us vtadmin-api --> vtgate_ap vtadmin-api --> vtctld_ap subgraph us-east-1 topology_us["Topology Service"] vtgate_us["vtgate"] vtctld_us["vtctld"] vtgate_us <--> topology_us vtctld_us <--> topology_us end subgraph ap-east-1 topology_ap["Topology Service"] vtgate_ap["vtgate"] vtctld_ap["vtctld"] vtgate_ap <--> topology_ap vtctld_ap <--> topology_ap end

The Life Cycle of a VTAdmin Request #

To understand how data moves through VTAdmin, let's look at life cycle of a typical request: fetching a list of all of the schemas. We'll use the single-cluster "Get Started" environment as an example:

A list of schemas in VTAdmin

When a user loads the /schemas view in the browser, vtadmin-web makes an HTTP GET /api/schema/local/commerce/corder request to vtadmin-api. vtadmin-api then issues gRPC requests to the vtgates and vtctlds in the cluster to construct the list of schemas. Here's what that looks like in detail:

sequenceDiagram autonumber participant W as vtadmin-web participant A as vtadmin-api participant G as vtgate participant V as vtctld W ->> A: GET /api/schemas A ->> G: SHOW vitess_tablets A ->> V: GetKeyspaces A ->> V: GetSchema Note left of V: { Keyspace: "commerce", TabletAlias: "zone1-101" } A ->> V: GetSchema Note left of V: { Keyspace: "customer", TabletAlias: "zone1-201" } A ->> W: 200 OK Note right of W: { "schemas": [...] }
  1. vtadmin-web makes a GET /api/schema/local/commerce/corder request against vtadmin-api's HTTP endpoint.
  2. vtadmin-api discovers a vtgate in the cluster and issues a SHOW vitess_tablets query on that vtgate to get a list of all tablets in the cluster.
  3. vtadmin-api discovers a vtctld in the cluster and makes a GetKeyspaces gRPC request to get a list of all keyspaces in the cluster.
  4. For each of these keyspaces, vtadmin-api chooses a random, serving tablet. The keyspace and tablet alias are used to make a GetSchema gRPC request to the vtctld to get the full schema, with all of the tables, on that tablet in the keyspace.
  5. Since the "Get Started" example has two keyspaces, commerce and customer, vtadmin-api issues separate requests in parallel.
  6. Finally, the GetSchema gRPC response is annotated with additional metadata (like cluster information), and returned to vtadmin-web as JSON.
Architecture