Server Crash Due to Full Root File System?

[jjk3]

I have been working with Solaris for about 5 years now and I have yet to see a server crash because it's root file system is 100% full. I have even done experiments where I purposely fill the root file system just to try and get the system to crash. All of my servers have been partitioned with at least /var as a separate partition.

I have seen many sources quote that a server will crash if the root file system is full. Is this true or urban myth?

Has anyone seen a server crash because of a full root file system?

Thanks

Joe


---------------------------------------
Joe Keegan - Joe@jjk3.com
SANS GSEC & GCFW
CCSE, CCNA, CCSA & Sun Certified

 

[AHinMaine]

It's true but it's difficult to accomplish because I believe the default is to reserve 3% of the file system. Don't have my book handy or I'd look it up for ya. It's in the SysAdminI book. --
Andy

http://www.nachoz.com

 

[jjk3]

I remember that there was some space reserved (3 or 5%), but I thought root could write to this space. I have written a script which simply dumps garbage into a file in the root file system until the script fails because the device is full. I run the script as root so I assume that I have privileges to really fill up the file system.

The more important question is why? Why would a server crash if the root file system was full?

Thanks for helping answer my curiosity.

Joe
---------------------------------------
Joe Keegan - Joe@jjk3.com
SANS GSEC & GCFW
CCSE, CCNA, CCSA & Sun Certified

 

[AllenV]

Typically - my experience has been that /var - not being it's own filespace will fill up and crash the system. If /var is not it's own filesystem, but part of root - your utmp will be unwritable. THAT will crash a system fast.