Cent

======

Possible backwards incompatibility here (in client side code) - see first point.

* omit fields in message JSON if field contains empty value: `client` on top level, `info` on top level, `default_info` in `info` object, `channel_info` in `info` object. This also affects top level data in join/leave messages and presence data – i.e. `default_info` and `channel_info` keys not included in JSON when empty. This can require adapting your client side code a bit if you rely on these keys but for most cases this should not affect your application. But we strongly suggest to test before updating. This change allows to reduce message size. See migration notes below for more details.
* new option `--admin_port` to bind admin websocket and web interface to separate port. [44](https://github.com/centrifugal/centrifugo/issues/44)
* new option `--api_port` to bind API endpoint to separate port. [44](https://github.com/centrifugal/centrifugo/issues/44)
* new option `--insecure_web` to use web interface without setting `web_password` and `web_secret` (for use in development or when you protected web interface by firewall rules). [44](https://github.com/centrifugal/centrifugo/issues/44)
* new channel option `history_drop_inactive` to drastically reduce resource usage (engine memory, messages travelling around) when you use message history. See [50](https://github.com/centrifugal/centrifugo/issues/50)
* new Redis engine option `--redis_api_num_shards`. This option sets a number of Redis shard queues Centrifugo will use in addition to standard `centrifugo.api` queue. This allows to increase amount of messages you can publish into Centrifugo and preserve message order in channels. See [52](https://github.com/centrifugal/centrifugo/issues/52) and [documentation](https://fzambia.gitbooks.io/centrifugal/content/server/engines.html) for more details.
* fix race condition resulting in client disconnections on high channel subscribe/unsubscribe rate. [54](https://github.com/centrifugal/centrifugo/issues/54)
* refactor `last_event_id` related stuff to prevent memory leaks on large amount of channels. [48](https://github.com/centrifugal/centrifugo/issues/48)
* send special disconnect message to client when we don't want it to reconnect to Centrifugo (at moment to client sending malformed message).
* pong wait handler for raw websocket to detect non responding clients.

Also it's recommended to update javascipt client to latest version as it has some useful changes (see its changelog).

How to migrate
--------------

Message before:

json
{
"uid":"442586d4-688c-4a0d-52ad-d0a13d201dfc",
"timestamp":"1450817253",
"info": null,
"channel":"$public:chat",
"data":{"input":"1"},
"client":""
}


Message now:

json
{
"uid":"442586d4-688c-4a0d-52ad-d0a13d201dfc",
"timestamp":"1450817253",
"channel":"$public:chat",
"data":{"input":"1"}
}


I.e. not using empty `client` and `info` keys. If those keys are non empty then they present in message.

Join message before:

json
{
"user":"2694",
"client":"93615872-4e45-4da2-4733-55c955133436",
"default_info": null,
"channel_info":null
}


Join message now:

json
{
"user":"2694",
"client":"93615872-4e45-4da2-4733-55c955133436"
}


If "default_info" or "channel_info" exist then they would be included:

json
{
"user":"2694",
"client":"93615872-4e45-4da2-4733-55c955133436",
"default_info": {"username": "FZambia"},
"channel_info": {"extra": "some data here"}
}

======

No backwards incompatible changes here.

* New `recover` option to automatically recover missed messages based on last message ID. See [pull request](https://github.com/centrifugal/centrifugo/pull/42) and [chapter in docs](https://fzambia.gitbooks.io/centrifugal/content/server/recover.html) for more information. Note that you need centrifuge-js >= v1.1.0 to use new `recover` option
* New `broadcast` API method to send the same data into many channels. See [issue](https://github.com/centrifugal/centrifugo/issues/41) and updated [API description in docs](https://fzambia.gitbooks.io/centrifugal/content/server/api.html)
* Dockerfile now checks SHA256 sum when downloading release zip archive.

======

No backwards incompatible changes here.

* support enabling web interface over environment variable CENTRIFUGO_WEB
* close client's connection after its message queue exceeds 10MB (default, can be modified using `max_client_queue_size` configuration file option)
* fix theoretical server crash on start when reading from redis API queue

======

A bad and a good news here. Let's start with a good one. Centrifugo is still real-time messaging server and just got v1.0 release. The bad – it is not fully backwards compatible with previous versions. Actually there are three changes that ruin compatibility. If you don't use web interface and private channels then there is only one change. But it affects all stack - Centrifugo itself, client library and API library.

Starting from this release Centrifugo won't support multiple registered projects. It will work with only one application. You don't need to use `project key` anymore. Changes resulted in simplified
configuration file format. The only required option is `secret` now. See updated documentation
to see how to set `secret`. Also changes opened a way for exporting Centrifugo node statistics via HTTP API `stats` command.

As this is v1 release we'll try to be more careful about backwards compatibility in future. But as we are trying to make a good software required changes will be done if needed.

Highlights of this release are:

* Centrifugo now works with single project only. No more `project key`. `secret` the only required configuration option.
* web interface is now embedded, this means that when downloading release you get possibility to run Centrifugo web interface just providing `--web` flag to `centrifugo` when starting process.
* when `secret` set via environment variable `CENTRIFUGO_SECRET` then configuration file is not required anymore. But note, that when Centrifugo configured via environment variables it's not possible to reload configuration sending HUP signal to process.
* new `stats` command added to export various node stats and metrics via HTTP API call. Look its response example [in docs chapter](https://fzambia.gitbooks.io/centrifugal/content/server/api.html).
* new `insecure_api` option to turn on insecure HTTP API mode. Read more [in docs chapter](https://fzambia.gitbooks.io/centrifugal/content/mixed/insecure_mode.html).
* minor clean-ups in client protocol. But as protocol incapsulated in javascript client library you only need to update centrifuge-js.

======

* new `channels` API command – allows to get list of active channnels in project at moment (with one or more subscribers).
* `message_send_timeout` option default value is now 0 (last default value was 60 seconds) i.e. send timeout is not used by default. This means that Centrifugo won't start lots of goroutines and timers for every message sent to client. This helps to drastically reduce memory allocations. But in this case it's recommended to keep Centrifugo behind properly configured reverse proxy like Nginx to deal with connection edge cases - slow reads, slow writes etc.
* Centrifugo now sends pings into pure Websocket connections. Default value is 25 seconds and can be adjusted using `ping_interval` configuration option. Note that this option also sets SockJS heartbeat messages interval. This opens a road to set reasonable value for Nginx `proxy_read_timeout` for `/connection` location to mimic behaviour of `message_send_timeout` which is now not used by default
* improvements in Redis Engine locking.
* tests now require Redis instance running to test Redis engine. Tests use Redis database 9 to run commands and if that database is not empty then tests will fail to prevent corrupting existing data.
* all dependencies now vendored.

======

* HTTP API endpoint now can handle json requests. Used in client written in Go at moment. Old behaviour have not changed, so this is absolutely optional.
* dependency packages updated to latest versions - websocket, securecookie, sockjs-go.