Skip to content

Latest commit

 

History

History
253 lines (185 loc) · 10 KB

README.md

File metadata and controls

253 lines (185 loc) · 10 KB
Raw

Logplex

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/.

Table of Contents

  1. Build
  2. Test
  3. Develop
  4. Supervision Tree
  5. Processes

Build

$ ./rebar --config public.rebar.config get-deps compile

Test

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.

Develop

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

Supervision Tree

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

Processes

logplex_db

Starts and supervises a number of ETS tables:

channels
tokens
drains
creds
sessions

config_redis

A redo redis client process connected to the logplex config redis.

logplex_drain_sup

An empty one_for_one supervisor. Supervises HTTP and TCP drain processes.

nsync

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.

redgrid

A redgrid process that registers the node in a central redis server to facilitate discovery by other nodes.

logplex_realtime

Captures realtime metrics about the running logplex node. This metrics are exported using folsom_cowboy and are available for consumption via HTTP.

Memory Usage information is available:

> curl -s http://localhost:5565/_memory | jq '.'
{
  "total": 27555464,
  "processes": 10818248,
  "processes_used": 10818136,
  "system": 16737216,
  "atom": 388601,
  "atom_used": 371948,
  "binary": 789144,
  "code": 9968116,
  "ets": 789128
}

As is general VM statistics:

> curl -s http://localhost:5565/_statistics | jq '.'
{
  "context_switches": 40237,
  "garbage_collection": {
    "number_of_gcs": 7676,
    "words_reclaimed": 20085443
  },
  "io": {
    "input": 9683207,
    "output": 2427112
  },
  "reductions": {
    "total_reductions": 6584440,
    "reductions_since_last_call": 6584440
  },
  "run_queue": 0,
  "runtime": {
    "total_run_time": 1140,
    "time_since_last_call": 1140
  },
  "wall_clock": {
    "total_wall_clock_time": 207960,
    "wall_clock_time_since_last_call": 207748
  }
}

Several custom logplex metrics are also exported via a special /_metrics endpoint:

> curl -s http://localhost:5565/_metrics | jq '.'
[
  "drain.delivered",
  "drain.dropped",
  "message.processed",
  "message.received"
]

These can then be queried individually:

> curl -s http://localhost:5565/_metrics/message.received | jq '.'
{
  "type": "gauge",
  "value": 1396
}

logplex_stats

Owns the logplex_stats ETS table. Prints channel, drain and system stats every 60 seconds.

logplex_tail

Maintains the logplex_tail ETS table that is used to register tail sessions.

logplex_redis_writer_sup

Starts a logplex_worker_sup process, registered as logplex_redis_writer_sup, that supervises logplex_redis_writer processes.

logplex_read_queue_sup

Starts a logplex_queue_sup process, registered as logplex_read_queue_sup, that supervises logplex_queue processes.

logplex_reader_sup

Appears to start a logplex_worker_sup that supervises logplex_reader processes. That model doesn't seem to exist?!?

logplex_shard

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.

logplex_api

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.

logplex_syslog_sup

Supervises a tcp_proxy_sup process that supervises a tcp_proxy process that accepts syslog messages over TCP.

logplex_logs_rest

Starts a cowboy_tcp_transport process and serves as the callback for processing HTTP log input.

Realtime Metrics

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.