Well-used services write reams of log information to disk, which is not only bothersome from a storage perspective but also pushes grep and the usual group of statistics tools to their limits. Will hitching the syslog daemon up to a database help?
Who said length doesn’t matter? My spam filters alone give me a 3GB logfile daily, which would be fine if I just needed the beast to check up on the occasional error. Because I need to extract a whole bunch of statistics about spam and virus threats from the file, grepping such enormous files takes ages, besides creating far too much I/ O overhead.
RSyslog took me a giant leap closerm to finding a solution – it logs directly to a MySQL or PostgreSQL database, meaning that I can replace my grep commands with fast SQL statements. RSyslog is included with many distributions and is the default application in Fedora 8, for example. By default, my Ubuntu lab environment runs sysklogd, making it easy for me to switch to RSyslog. For the time being, I can even keep my old syslog.conf. For the most part, RSyslog’s configuration file is the same as the legacy format, although it does support a couple of additional options for linking up with the database.
Watch 3 days full of Apache talks live from Amsterdam on March 25-27 in the convenience of your home or office. Topics are: Apache Hadoop, Tomcat for Developers and Administrators, HTTP Server Administration and much more.
Comments