Skip to content
Find file
Fetching contributors…
Cannot retrieve contributors at this time
64 lines (50 sloc) 2.84 KB
dklab_logreplica: gathers logs from multiple machines into one place in realtime
(C) Dmitry Koterov,
License: GPL
This simple, but very handy and powerful tool allows you to pull logs
from the whole cluster to a single logging server in realtime. These
logs may be used for monitoring, aggregation or analyzing.
1. Choose your logging server: it is a machine with root account and
/opt/dklab_logreplica/ directory holding dklab_logreplica scripts.
2. Copy dklab_logreplica.init into /etc/init.d/dklab_logreplica and
setup its automatical startup at server boot time.
3. Copy default config file dklab_logreplica.conf.sample to
/etc/dklab_logreplica.conf and modify it according to your needs:
- specify which host you want to pull the logs from,
- what files to replicate (you may specify wildcards freely)
- what destination directory is used to put the logs to.
4. Create SSH private-public key pair using "ssh-keygen -t rsa".
Put public key to each of logs source machines using
"ssh-copy-id root@machine-to-be-pulled" - you should have
access to all these machines from the log server without
entering a password.
5. Now run /etc/init.d/dklab_logreplica start
After all that steps logs from source machines will be automatically
and continuosly replicated to the log server in the background (with
support of reconnects and resumes if the connection is not stable).
You may monitor the activity at standard /var/log/messages file.
No need to configure source machines - the single configuration
point is dklab_logreplica.conf at your logging server.
If you have many machines in your cluster which perform different tasks
(e.g. SQL server, web-frontend, balancer, mail server etc.), you may want
to aggregate logs from all these machines to a single place to monitor them
or to preform various statistics collections. Of course you may configure
syslog or syslog-ng to pass all the data over the network, but if you do
so, you are face to face with the following problems:
1. Due to network problems pieces of logs may not be correctly transfered,
so you may loose data.
2. It is quite hard to keep the syslog configuration in sync with real
world (which is changing time to time, you know).
3. Not all services support putting logs into syslog (e.g. apache supports
only file-based logs writing). So you have to use named pipes, and your
configs grow.
4. Quite oftenly you want to replicate logs filenames by their wildcards
specified. Syslog cannot do that.
5. At least, it is good to hold all logs at the machines at which they
are produced (with e.g. weekly rotation) in addition to send them
to a logging server. So your configs are growing again...
Dklab_logreplica solves all that problems.
Something went wrong with that request. Please try again.