tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: LFS without lfs_cleanerd?



On Mon, 24 Mar 2008, J.T. Conklin wrote:

What would be the expected failure mode of a LFS filesystem wasn't
running the lfs_cleaner daemon?  Due to a packaging error, I had a
wrong version of lfs_cleanerd in a NetBSD-based appliance firmware
image. After some months, all writes to the filesystem blocked.

Yes, that would be the expected result if there is no lfs_cleanerd running. The kernel automatically cleans empty segments, so you might not notice it for a while, but at some point you will likely need to clean to be able to write.

operations) in a loop for about a week without being able to reproduce
the failure. Is there any specific set of operations that would make a
better accelerated wear test?

Try something like this. Fill the filesystem to, say, 80% with small files, written sequentially (file0, file1, file2, ...; "small" means "much smaller than the segment size", maybe 1 block per file). Then delete and rewrite the odd-numbered ones. Most segments should be around half-full after the delete, so they won't be able to be reused without cleaning. Without segment reuse the filesystem would end up 120% full; it should block instead.

Take care,

-----------------------------------------------------------------------
Konrad Schroder               Box 352195 -or- 14 Loew Hall, Stevens Way
Information Tech & Services                    University of Washington
Human Interface Technology Lab                  Seattle, WA, 98195, USA


Home | Main Index | Thread Index | Old Index