Skip to content

Latest commit

 

History

History
97 lines (69 loc) · 5.55 KB

INSTALL.md

File metadata and controls

97 lines (69 loc) · 5.55 KB

Installing Tinode

Using Docker

See instructions

Building from Source

  • Install Go environment

  • Install RethinkDB

  • Fetch, build tinode server and tinode-db database initializer:

  • go get github.com/tinode/chat/server && go install -tags rethinkdb github.com/tinode/chat/server

  • go get github.com/tinode/chat/tinode-db && go install -tags rethinkdb github.com/tinode/chat/tinode-db

Note the required -tags rethinkdb build option. You may also optionally define main.buildstamp for the server by adding a build option

-X main.buildstamp=`date -u '+%Y%m%dT%H:%M:%SZ'`

This build timestamp will be sent by the server to the clients.

Running standalone server

  • Run RethinkDB: rethinkdb --bind all --daemon

  • Run DB initializer $GOPATH/bin/tinode-db -config=$GOPATH/src/github.com/tinode/chat/server/tinode.conf

  • add -data=$GOPATH/src/github.com/tinode/chat/tinode-db/data.json flag if you want sample data to be loaded.

DB intializer needs to be run only once per installation. See instructions for more options.

Running a cluster

  • Install RethinkDB, run it stanalone or in cluster mode. Run DB initializer, unpack JS files as described in the previous section.

  • Cluster expects at least two nodes. A minimum of three nodes is recommended. A sample cluster config file is provided as cluster.conf. The following section configures the cluster.

	"cluster_config": {
		"nodes": [
			{"name": "one", "addr":"localhost:12001"},
			{"name": "two", "addr":"localhost:12002"},
			{"name": "three", "addr":"localhost:12003"}
		],
		"self": "one",
		"failover": {
			"enabled": true,
			"heartbeat": 100,
			"vote_after": 8,
			"node_fail_after": 16
		}
	}
  • nodes defines individual cluster nodes. The sample defines three nodes named one, two, and tree running at the localhost at the specified cluster communication ports. Cluster addresses don't need to be exposed to the clients.
  • self is the name of the current node. Generally it's more convenient to specify the name of the current node at the command line using cluster_self option. Command line value overrides the config file value.
  • failover is an experimental feature which migrates topics from failed cluster nodes keeping them accessible:
    • enabled turns on failover mode; failover mode requires at least three nodes in the cluster.
    • heartbeat interval in milliseconds between heartbeats sent by the leader node to follower nodes to ensure they are accessible.
    • vote_after number of failed heartbeats before a new leader node is elected.
    • node_fail_after number of heartbeats that a follower node misses before it's cosidered to be down.

If you are testing the cluster with all nodes running on the same host, you also must override the listen port. Here is an example for launching two cluster nodes from the same host using the same config file:

./server -config=./cluster.conf -static_data=./example-react-js/ -listen=:6060 -cluster_self=one &
./server -config=./cluster.conf -static_data=./example-react-js/ -listen=:6061 -cluster_self=two &

Note on running the server in background

There is no clean way to daemonize a Go process internally. One must use external tools such as shell & operator, systemd, launchd, SMF, daemon tools, runit, etc. to run the process in the background.

Specific note for nohup users: an exit must be issued immediately after nohup call to close the foreground session cleanly:

> nohup $GOPATH/bin/server -config=$GOPATH/src/github.com/tinode/chat/server/tinode.conf -static_data=$HOME/tinode/example-react-js/ &
> exit

Otherwise SIGHUP may be received by the server if the shell connection is broken before the ssh session has terminated (indicated by Connection to XXX.XXX.XXX.XXX port 22: Broken pipe). In such a case the server will shutdown because SIGHUP is intercepted by the server and interpreted as a shutdown request.

For more details see tinode#25.