http://oss.sgi.com/bugzilla/show_bug.cgi?id=209
------- Additional Comments From tdickson@xxxxxxxxxxx 2003-10-12 08:27 PDT
-------
That sounds like it may be usefull.
When I say xfs_check dies I mean that it uses memory until there is none left,
and then the oom killer kills it; resulting in this error message:
Also, I know dir creation at the same time that
tree
is being run is similar to a fork-bomb, but I'm trying to recreate the scenario
on a test machine. On the production machine, a simple "tree" or rsync command
(anything that reads all of the inodes (or a large number) causes the memory to
be used until the system is EFFECTIVLY crashed (it doesn't actually die, and
comes back later, but it kills httpd/samba/atalk/etc in the process).
Mihai suggestions look interesting; I'll try with the 2.4.23 kernel and see if
removing the oom killer helps. To me, also, it looks as if a cache is too
aggressive, given that the machine eventually let go of the xfs_inode memory.
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
|