Skip to content

Latest commit

 

History

History
64 lines (38 loc) · 3.45 KB

GettingStarted3.md

File metadata and controls

64 lines (38 loc) · 3.45 KB

Getting started - Three node cluster

In the Getting started page, we setup a single node cluster. Here, we setup a three node node cluster, which is tolerant to a single node failure. We call this setup highly available (HA).

Setup cluster

Following is a step-by-step guide to getting this HA cluster up and running. This setup has three cluster node processes and one beanstalkd proxy process. You can see a more production-like setup for kubernetes here

Setup cluster-node bean0

# create a directory for demonstration purposes.
$ mkdir -p /tmp/ha

$ ./coolbeans --quiet cluster-node --node-id bean0 --root-dir /tmp/ha/bean0 --bootstrap-node-id bean0 --node-peer-addrs=:11000,:12000,:13000 --raft-listen-addr=127.0.0.1:21000 --node-listen-addr=127.0.0.1:11000 --raft-advertized-addr=127.0.0.1:21000 --prometheus-addr=127.0.0.1:2020

The above example starts a new cluster node

  • with a unique node-id bean0.
  • the cluster-node starts two listeners, the raft service listens on 127.0.0.1:21000 and the GRPC service listens on :11000.
  • the raft advertised address indicates the address the other raft peers will use to connect.
  • all data logs and snapshots will be persisted under the directory: /tmp/ha/bean0.
  • the bootstrap node, which the default assigned leader during the first time cluster formation, is bean0.
  • additionally, we define all the initial peers: :11000,:12000,:13000.

Setup cluster-nodes bean1 & bean2

./coolbeans --quiet cluster-node --node-id bean1 --root-dir /tmp/ha/bean1 --bootstrap-node-id bean0 --node-peer-addrs=:11000,:12000,:13000 --raft-listen-addr=127.0.0.1:22000 --node-listen-addr=127.0.0.1:12000 --raft-advertized-addr=127.0.0.1:22000 --prometheus-addr=127.0.0.1:2021

./coolbeans --quiet cluster-node --node-id bean2 --root-dir /tmp/ha/bean2 --bootstrap-node-id bean0 --node-peer-addrs=:11000,:12000,:13000 --raft-listen-addr=127.0.0.1:23000 --node-listen-addr=127.0.0.1:13000 --raft-advertized-addr=127.0.0.1:23000 --prometheus-addr=127.0.0.1:2022

The above example starts two cluster nodes

  • with unique ids bean1 and bean2 which join node bean0 to form a three node cluster.
  • the bootstrap node in this case is still bean0. What that means, is if we bring up all the three nodes in parallel, only bean0 becomes the leader, and bean1 and bean2 become followers.

Setup beanstalkd proxy

 ./coolbeans --quiet beanstalkd --listen-addr 127.0.0.1 --upstream-addrs 127.0.0.1:11000,127.0.0.1:12000,127.0.0.1:13000 --listen-port 11300 >> /tmp/coolbeans-sidecar.log 2>> /tmp/coolbeans-sidecar_error.log

The above example starts a new beanstalkd proxy:

  • with the proxy upstream pointing to the GRPC services: 127.0.0.1:11000, 127.0.0.1:12000 and 127.0.0.1:13000. The proxy automatically detects which of three is the leader and forwards all the requests to the current leader.
  • listening for beanstalkd client requests on port 11300.

Query to find out which node is the elected leader

$  ./coolbeans cluster-client is_leader --node-addr 127.0.0.1:11000
isNodeLeader: false

$ ./coolbeans cluster-client is_leader --node-addr 127.0.0.1:12000
isNodeLeader: false

$ ./coolbeans cluster-client is_leader --node-addr 127.0.0.1:13000
isNodeLeader: true

From the above, you can see that the node bean2 was elected the leader even though the boostrapped leader initially was bean0.