Find is one of my favorite little tools under linux. It helps me “find” almost anything, I can find files older than a certain date, newer than a certain date, modified on a certain date. I can find files that have a certain name, or match a part of a name, file extension. Once I’ve found what I’ve been looking for I can have find do something with those files like delete them or gzip them.
My latest “find” with the find command came about because on one of my JBoss servers I wrote a simple script that looks for log files older than 15 days and deletes them and looks for other log files older than 61 minutes and compresses them with gzip.
#!/bin/bash
LOGS=/usr/local/jboss/server/all/log/
#delete all logs older than 37 days
find $LOGS -mtime +15 | xargs rm -rf
# gzip files last modify at least 1 hour ago
find $LOGS -mmin +61 | xargs gzip
Our JBoss setup automatically writes new logs to server.info.log and server.error.log, then every every hour it renames the INFO and ERROR log to the current date + hour, so server.info.log would be changed to server.info.log.2010-03-09-13 for today at 2pm to roll out the 1pm logs.
The problem I came across in my script was with my server.error.log file. If an error hasn’t been written to the server.error.log file during that hour, it wasn’t going to rotate an empty error log. Since the file hadn’t been touched/updated/modified in over 61 minutes, my script came along and gzipped it, at this point JBoss then had a problem because the error log was missing and didn’t create a new one.
So what I needed to do was to find all the files that matched the criteria, but exclude the server.info.log and server.error.log and here is my final script.
#!/bin/bash
LOGS=/usr/local/jboss/server/all/log/
INFOLOG=”server.info.log”
ERRORLOG=”server.error.log”
#delete all logs older than 37 days
find $LOGS -mtime +15 -not -name “$INFOLOG” -not -name “$ERRORLOG” | xargs rm -rf
# gzip files last modify at least 1 hour ago
find $LOGS -mmin +61 -not -name “$INFOLOG” -not -name “$ERRORLOG” | xargs gzip
