Solaris

Hi Experts,

I had encountered an issue where the zero link process was holding too much amount of data on a Solaris server. I was able to terminate the process after which the space of the file system was released.


40G /proc/8567/fd/2

Can you please let me know if there are any special patches released by Oracle to address this issue.
Thanks in advance.

Why do you think a patch is required ?

This was caused by a process that had its stderr redirected to a file that was later removed, more than likely by someone wrongly thinking the disk space was going to be released. This is the way Unix works.

To recover the space, instead of killing the process, you could have blanked the file instead with this command:

Code:

: >  /proc/8567/fd/2

Thank you for your response jilliagre.

Can you please let me know how do we prevent the "stderr" of the process from hanging and occupying the space this way.

One way would be to redirect stderr to /dev/null

A better way would be to figure out what the process is trying hard to tell you when writing forty gigabytes of error messages, then fix the root cause.

Don't use redirected stdout or stderr for logging. Use proper logging where the log file isn't tied to the running process.

Even proper logging will misbehave this way if you blithely delete the file while it's in use. It's a very common beginner administration mistake. Just truncate the file instead of deleting it.