» Consul Architecture
Consul is a complex system that has many different moving parts. To help users and developers of Consul form a mental model of how it works, this page documents the system architecture.
Advanced Topic! This page covers technical details of the internals of Consul. You don't need to know these details to effectively operate and use Consul. These details are documented here for those who wish to learn about them without having to go spelunking through the source code.
Before describing the architecture, we provide a glossary of terms to help clarify what is being discussed:
Agent - An agent is the long running daemon on every member of the Consul cluster. It is started by running
consul agent. The agent is able to run in either client or server mode. Since all nodes must be running an agent, it is simpler to refer to the node as being either a client or server, but there are other instances of the agent. All agents can run the DNS or HTTP interfaces, and are responsible for running checks and keeping services in sync.
Client - A client is an agent that forwards all RPCs to a server. The client is relatively stateless. The only background activity a client performs is taking part in the LAN gossip pool. This has a minimal resource overhead and consumes only a small amount of network bandwidth.
Server - A server is an agent with an expanded set of responsibilities including participating in the Raft quorum, maintaining cluster state, responding to RPC queries, exchanging WAN gossip with other datacenters, and forwarding queries to leaders or remote datacenters.
Datacenter - We define a datacenter to be a networking environment that is private, low latency, and high bandwidth. This excludes communication that would traverse the public internet, but for our purposes multiple availability zones within a single EC2 region would be considered part of a single datacenter.
Consensus - When used in our documentation we use consensus to mean agreement upon the elected leader as well as agreement on the ordering of transactions. Since these transactions are applied to a finite-state machine, our definition of consensus implies the consistency of a replicated state machine. Consensus is described in more detail on Wikipedia, and our implementation is described here.
Gossip - Consul is built on top of Serf which provides a full gossip protocol that is used for multiple purposes. Serf provides membership, failure detection, and event broadcast. Our use of these is described more in the gossip documentation. It is enough to know that gossip involves random node-to-node communication, primarily over UDP.
LAN Gossip - Refers to the LAN gossip pool which contains nodes that are all located on the same local area network or datacenter.
WAN Gossip - Refers to the WAN gossip pool which contains only servers. These servers are primarily located in different datacenters and typically communicate over the internet or wide area network.
RPC - Remote Procedure Call. This is a request / response mechanism allowing a client to make a request of a server.
» 10,000 foot view
From a 10,000 foot altitude the architecture of Consul looks like this:
Let's break down this image and describe each piece. First of all, we can see that there are two datacenters, labeled "one" and "two". Consul has first class support for multiple datacenters and expects this to be the common case.
Within each datacenter, we have a mixture of clients and servers. It is expected that there be between three to five servers. This strikes a balance between availability in the case of failure and performance, as consensus gets progressively slower as more machines are added. However, there is no limit to the number of clients, and they can easily scale into the thousands or tens of thousands.
All the nodes that are in a datacenter participate in a gossip protocol. This means there is a gossip pool that contains all the nodes for a given datacenter. This serves a few purposes: first, there is no need to configure clients with the addresses of servers; discovery is done automatically. Second, the work of detecting node failures is not placed on the servers but is distributed. This makes failure detection much more scalable than naive heartbeating schemes. Thirdly, it is used as a messaging layer to notify when important events such as leader election take place.
The servers in each datacenter are all part of a single Raft peer set. This means that they work together to elect a single leader, a selected server which has extra duties. The leader is responsible for processing all queries and transactions. Transactions must also be replicated to all peers as part of the consensus protocol. Because of this requirement, when a non-leader server receives an RPC request, it forwards it to the cluster leader.
The server nodes also operate as part of a WAN gossip pool. This pool is different from the LAN pool as it is optimized for the higher latency of the internet and is expected to contain only other Consul server nodes. The purpose of this pool is to allow datacenters to discover each other in a low-touch manner. Bringing a new datacenter online is as easy as joining the existing WAN gossip pool. Because the servers are all operating in this pool, it also enables cross-datacenter requests. When a server receives a request for a different datacenter, it forwards it to a random server in the correct datacenter. That server may then forward to the local leader.
This results in a very low coupling between datacenters, but because of failure detection, connection caching and multiplexing, cross-datacenter requests are relatively fast and reliable.
In general, data is not replicated between different Consul datacenters. When a request is made for a resource in another datacenter, the local Consul servers forward an RPC request to the remote Consul servers for that resource and return the results. If the remote datacenter is not available, then those resources will also not be available, but that won't otherwise affect the local datacenter. There are some special situations where a limited subset of data can be replicated, such as with Consul's built-in ACL replication capability, or external tools like consul-replicate.
In some places, client agents may cache data from the servers to make it available locally for performance and reliability. Examples include Connect certificates and intentions which allow the client agent to make local decisions about inbound connection requests without a round trip to the servers. Some API endpoints also support optional result caching. This helps reliability because the local agent can continue to respond to some queries like service-discovery or Connect authorization from cache even if the connection to the servers is disrupted or the servers are temporarily unavailable.
» Getting in depth
At this point we've covered the high level architecture of Consul, but there are many more details for each of the subsystems. The consensus protocol is documented in detail as is the gossip protocol. The documentation for the security model and protocols used are also available.
For other details, either consult the code, ask in IRC, or reach out to the mailing list.