Logplex is a distributed syslog log router, able to merge and redistribute multiple incoming streams of syslog logs to individual subscribers.
A typical logplex installation will be a cluster of distributed Erlang nodes connected in a mesh, with one or more redis instances (which can be sharded). The cluster may or may not be sitting behind a load-balancer or proxy, but any of them may be contacted at any time for ideal scenarios.
Applications sitting on their own node or server need to send their log messages either to a local syslog, or though log shuttle, which will then forward them to one instance of a logplex router.
On the other end of the spectrum, consumers may subscribe to a logplex instance, which will then merge streams of incoming log messages and forward them to the subscriber. Alternatively, the consumer may register a given endpoint (say, a database behind the proper API) and logplex nodes will be able to push messages to that end-point as they come in.
For more details, you can look at stream management documentation in doc/.
$ ./rebar --config public.rebar.config get-deps compile
Given an empty local redis (v2.6ish):
$ ./rebar get-deps compile --config test.rebar.config
$ INSTANCE_NAME=`hostname` \
LOGPLEX_CONFIG_REDIS_URL="redis://localhost:6379" \
LOGPLEX_SHARD_URLS="redis://localhost:6379" \
LOGPLEX_REDGRID_REDIS_URL="redis://localhost:6379" \
LOCAL_IP="127.0.0.1" \
LOGPLEX_COOKIE=123 \
ERL_LIBS=`pwd`/deps/:$ERL_LIBS \
ct_run -spec logplex.spec -pa ebin
Runs the common test suite for logplex.
run
$ INSTANCE_NAME=`hostname` \
LOGPLEX_CONFIG_REDIS_URL="redis://localhost:6379" \
LOGPLEX_REDGRID_REDIS_URL="redis://localhost:6379" \
LOCAL_IP="127.0.0.1" \
LOGPLEX_COOKIE=123 \
LOGPLEX_AUTH_KEY=123 \
erl -name logplex@`hostname` -pa ebin -env ERL_LIBS deps -s logplex_app -setcookie ${LOGPLEX_COOKIE} -config sys
create creds
1> logplex_cred:store(logplex_cred:grant('full_api', logplex_cred:grant('any_channel', logplex_cred:rename(<<"Local-Test">>, logplex_cred:new(<<"local">>, <<"password">>))))).
ok
hit healthcheck
$ curl http://local:password@localhost:8001/healthcheck
OK
create a channel
$ curl -d '{"tokens": ["app"]}' http://local:password@localhost:8001/channels
{"channel_id":1,"tokens":{"app":"t.feff49f1-4d55-4c9e-aee1-2d2b10e69b42"}}
post a log msg
$ curl -v \
-H "Content-Type: application/logplex-1" \
-H "Logplex-Msg-Count: 1" \
-d "116 <134>1 2012-12-10T03:00:48.123456Z erlang t.feff49f1-4d55-4c9e-aee1-2d2b10e69b42 console.1 - - Logsplat test message 1" \
http://local:password@localhost:8601/logs
create a log session
$ curl -d '{"channel_id": "1"}' http://local:password@localhost:8001/v2/sessions
{"url":"/sessions/9d53bf70-7964-4429-a589-aaa4df86fead"}
fetch logs for session
$ curl http://local:password@localhost:8001/sessions/9d53bf70-7964-4429-a589-aaa4df86fead?srv=1
2012-12-10T03:00:48Z+00:00 app[console.1]: test message 1
| logplex_app | logplex_sup | logplex_db | ||
| config_redis (redo) | ||||
| logplex_drain_sup | logplex_http_drain | |||
| logplex_tcpsyslog_drain | ||||
| nsync | ||||
| redgrid | ||||
| logplex_realtime | redo | |||
| logplex_stats | ||||
| logplex_tail | ||||
| logplex_redis_writer_sup (logplex_worker_sup) | logplex_redis_writer | |||
| logplex_read_queue_sup (logplex_queue_sup) | logplex_queue | |||
| logplex_reader_sup (logplex_worker_sup) | logplex_reader | |||
| logplex_shard | redo | |||
| logplex_api | ||||
| logplex_syslog_sup | tcp_proxy_sup | tcp_proxy | ||
| logplex_logs_rest |
Starts and supervises a number of ETS tables:
channels
tokens
drains
creds
sessions
A redo redis client process connected to the logplex config redis.
An empty one_for_one supervisor. Supervises HTTP and TCP drain processes.
An nsync process connected to the logplex config redis. Callback module is nsync_callback.
Nsync is an Erlang redis replication client. It allows the logplex node to act as a redis slave and sync the logplex config redis data into memory.
A redgrid process that registers the node in a central redis server to facilitate discovery by other nodes.
Establishes a connection to the logplex stats redis. Owns the logplex_realtime ETS table and flushes the contents to tempo via redis pub/sub every second.
Metrics published to tempo:
message_received
message_processed
drain_delivered
drain_dropped
git_branch
availability_zone
Owns the logplex_stats ETS table. Prints channel, drain and system stats every 60 seconds.
Maintains the logplex_tail ETS table that is used to register tail sessions.
Starts a logplex_worker_sup process, registered as logplex_redis_writer_sup, that supervises logplex_redis_writer processes.
Starts a logplex_queue_sup process, registered as logplex_read_queue_sup, that supervises logplex_queue processes.
Appears to start a logplex_worker_sup that supervises logplex_reader processes. That model doesn't seem to exist?!?
Owns the logplex_shard_info ETS table. Starts a separate read and write redo client for each redis shard found in the logplex_shard_urls var.
Blocks waiting for nsync to finish replicating data into memory before starting a mochiweb acceptor that handles API requests for managing channels/tokens/drains/sessions.
Supervises a tcp_proxy_sup process that supervises a tcp_proxy process that accepts syslog messages over TCP.
Starts a cowboy_tcp_transport process and serves as the callback for processing HTTP log input.
Logplex can send realtime metrics to Redis via pubsub and to a drain channel as logs. The following metrics are currently logged in this fashion:
* `message_received`
* `message_processed`
* `drain_delivered`
* `drain_dropped`
To log these metrics to an internal drain channel, you'll need to set the
INTERNAL_METRICS_CHANNEL_ID environment variable to a drain token that has
already been created.