Thank you - found the culprit yesterday with du (did it before I got the post from Ward)
File called messages in var/log was over 10 GB!
It was not an accumulation of old messages as there were proper rotated copies from earlier days.
So it all must have been from Monday. Strange.
But there was an other strange thing Monday - maybe connected.
I noticed that the cpu bar was red at 100% (though the average was just showing over 1%)
I restarted the VM but after about 30 seconds the 100% came back though again the average showed only about 1.6%.
I closed down the VM and gave it 2 cores instead of 1 (could not think of anything else to try)
Restarted the VM and problem solved! Two hours after restart it all looked OK and the messages file has not yet been recreated since I deleted it.
So I restarted the VM with a single core and it was back at 100%
File messages was 25 MB after a few minutes!
The file is full of the following line (with only time changing):
Jan 26 20:22:12 pbxrand atd[26236]: File a000130169adf2 is in wrong format - aborting
I found the file in <var/spool/at> dated yesterday 10am with a partner of the same name but with an = instead of the a at the front of the name.
Both of length zero. Deleted both, shut down the VM, set it back to 1 core and now it all looks happy. Checked after 11 hours and all is well.
I generally do not look at this machine so it was most fortunate that I spotted to red bar, else I suppose the system would have been dead within an other few hours with a full disk.
Anyone know what that file is about and why the difference in behaviour between 1 and 2 cores?
As to the damaged file there is maybe a clue. The server on which the VM runs was force shut down yesterday morning so that is the most likely cause of the file damage.
I still had the disk showing at 75%. Needed a re-booting to bring it down to 24%.
Running PIAF Green-3.6.5.