Many years ago, I worked at Bolt, Beranek & Newman, and their unix machines had a utility called "roll". I suspect it was written by them for their machines, but it was a handy utility.

Basically, if you wanted to pipe output of a program into a file, but you didn't want that file to get massive, you'd do it something like this:

myprogram | roll -f log.out -size 500k

Once the file reached 500Kb, it would start overwriting itself, keeping only the most recent 500Kb of data.

Does something like this exist on standard Linux boxes today?

Hi,

what about using tail? There's the option --bytes=<size>

Regards

Thanks, but the problem is to keep large forgotten files from getting too large.

Doesn't the above keep log.out from exceeding 500Kb? Of course this doesn't prevent a _single_ output of yrprogram from temporarily making it exceed that size.

Sorry, I think I misunderstood what you were saying the first time around. I'll give that a try.

Thanks!

Don't worry. If you need to handle log files you might want to consider logrotate, which should be included in every distribution. It can handle files size as well.

Cheers