In one of my Solaris 8 boxes I have had this Oracle log file which was eating up my 2GB /u10/oracle partition. The file was 1.9GB in size. df -k was showing the partition was ~95% full. I have emptied the file by "bash# >/path/to/oracle.log" and that zeroed the file but the file size gone up again in 15-20 seconds to 1.9GB. Everytime I empty the file, the size goes back up again to 1.9GB in matter of a few seconds. But when I type df -k it doesn't show the partition is 95% full anymore. df -k shows the partition is 20% full. My question is: why df is not showing the disk is 95% full anymore and how come that file is filling up to 1.9GB very quickly?

This is a very primitive box and I can't use logrotate or anything like that. So I have a cronjob which empties the file every night.

Thanks.

The process writing to the file is likely seeking to former the log position. There is nothing you can do to avoid that. The log is then a sparse file so its reported size isn't matching the allocated blocks. This is standard behavior.