SUMMARY: tmpfs bytes/inode ratio?

From: David Luyer <>
Date: Thu May 09 2002 - 21:02:13 EDT
The unfortunately correct answer (from Matthew Stier, and actually
straight from the tmpfs manpage on Solaris 8, where I'd missed it):

      Another constraint is that the number of files available  in
      a   tmpfs  file  system  is calculated based on the physical
      memory of  the  machine   and  not  the  size  of  the  swap
      device/partition.   If  you  have too many files, tmpfs will
      print a warning message and you will be unable to create new
      files. You cannot increase this limit by adding swap space.

(ie. even though I only need a small amount of RAM for the CVS to run,
and a small amount of temp space, I need to increase the physical RAM
to let me have enough lock files/directories on the tmpfs)

There was one suggestion of tunefs, however that can only tune "real"
partitions (ones that have some form of device at least) and only
tunes contiguous blocks, rotational delay, block grouping, free
space precentage and space vs time optimization.

Some suggested tmpfs didn't use inodes as it was memory based.  This
isn't correct - there are still inodes, even if they might not be in
the same format as UFS inodes.  tmpfs supports hard links, persistance
of deleted files, all the common Unix file system semantics, so it
has a form of inodes.

As to working out the number of inodes on tmpfs, none of the standard,
xpg4 or ucb df's show it (in 'df -o i'), however GNU df (in 'df -i')
does accurately report the number of inodes used and free on tmpfs.
I don't know why the Solaris df's have this flaw.

David Luyer                                     Phone:   +61 3 9674 7525
Network Development Manager    P A C I F I C    Fax:     +61 3 9699 8693
Pacific Internet (Australia)  I N T E R N E T   Mobile:  +61 4 1111 BYTE                      NASDAQ:  PCNTF
sunmanagers mailing list
Received on Thu May 9 21:07:20 2002

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:42:42 EST