>
> I'm getting this sort of behaviour too. 1.3.80, there is disk activity
> every five or six seconds. procinfo says it's most often writing about
> three pages, but this varies. Under 1.2.13, on the same system, the
> closest I got to this was the minutely atrun. In both cases I have
> continuously running processes, including a top and three tail -fs. I
> suspect that the tails are what is causing this.
>
yep, "tail -f" reads (polls) the file every second, this in turn changes
the atime field, and bdflush updates the inode and the superblock.
might be a stupid question, but is there no way doing event based reads
on files [selects and/or blocking reads at the end of the file] ?
-- mingo