When you are sending lots of udp log events over the network packet loss can happen, or even using a tcp log sender you can get a slow response on your server depending on how much logs your remote log server is receiving simultaneously.
So... what can you do to avoid it?
logix can help you using its daemon receiving your log events and queueing you messages on AMQP. You can easily get rid of log event losses caused by udp and any performance issue that could be caused by concurrency using tcp remote syslog.
logix queues your log events on any AMQP Server and you can easy setup your graylog2-server to consume this queue and index your logs on demand.
# AMQP
amqp_enabled = true
amqp_subscribed_queues = logix:gelf
amqp_host = localhost
amqp_port = 5672
amqp_username = guest
amqp_password = guest
amqp_virtualhost = /
[transport]
connection_pool_enabled = False
connection_pool_size = 10
url = amqp://127.0.0.1:5672
queue = logix
[server]
port = 6660
max_syslog_line_size = 1023
bind_addr = 127.0.0.1
$ vim /etc/syslog.conf
*.notice;authpriv,remoteauth,ftp,install,internal.none @127.0.0.1:6660
$ launchctl unload /System/Library/LaunchDaemons/com.apple.syslogd.plist
$ launchctl load /System/Library/LaunchDaemons/com.apple.syslogd.plist
$ vim /etc/rsyslog.d/logix.conf
*.* @127.0.0.1:6660
$ /etc/init.d/rsyslog restart
$ Usage: ./logix
$ -h help
$ -u username
$ -d debug
$ -a <start|stop|status|foreground>
$ LOGIX_CONF=src/etc/logix.conf src/bin/logix -u $USER -a foreground -d &
$ logger test
- python-kombu (>= 1.4.3)
- python-gevent (>= 0.13.6)
- python-supay
- syslog or rsyslog :D
- would benefit of an internal backlog queue