log rotation based on month

[cdeardor]

I'm writing a script to rotate our syslogs. It used to be done by days, but now they want to keep 2 months online. So, for instance, if it's April, we'll archive February's logs. My problem is that I'm not sure how to manipulate the Month string from the date command, so that in January, and February, the script doesn't blow up. Any suggestions?

 

[SamBones]

You don't really need to manipulate a date string. You can use the [tt]find[/tt] command to select files based on their last modification or last access date and time.

Something like...

#!/bin/ksh

LOGDIR=/path/to/logs

find ${LOGDIR} -name '*.log' -mtime +60 -print | while read LOGFILE
do
    print "Log file to process: ${LOGFILE}"

    # do processing stuff here
done

This would find and allow you to process log files over 60 days since it was last modified. This works well over year end boundaries because it's based on the file's "age", not a date stamp of any kind.

You can also keep a file around longer if you want by just doing a "[tt]touch somefile.log[/tt]".

Hope this helps.

 

[cdeardor]

Thanks for the reply, however, that's how I originally had it done. Now, they don't want to go by 60 days, they want it keyed off of the month. So, 60 days won't work for me because of February only being 28 days, and also the 31 day months.

 

[hoinz]

Did you consider -newer option of find command?
Or to be more exact, its negation, ! -newer ?

 

[cdeardor]

That's a good idea, too. I definately didn't think of that one. I decided to just use case statements, and limit it to one of syslogs that can be kept online, after that, they get archived. I think I'll try some logic with the find command, though. Maybe that'll find it's way into the next version. Thanks! My big problem is that this is being done on AIX. For all the linux users out there, Here's a good command that would work with what I was trying to do:

date --date="2 month ago" +%b%d%Y

Thanks!