Date: Wed, 24 Oct 2001 01:11:27 -0700 (PDT) From: Matthew Dillon <dillon@apollo.backplane.com> To: Peter Pentchev <roam@ringlet.net> Cc: cvs-committers@FreeBSD.ORG, cvs-all@FreeBSD.ORG Subject: Re: cvs commit: src/sys/coda coda_subr.c src/sys/fs/msdosfs msdosfs_vfsops.c src/sys/fs/nwfs nwfs_vfsops.c src/sys/fs/smbfs smbfs_vfsops.c src/sys/fs/unionfs union_vfsops.c src/sys/gnu/ext2fs ext2_vfsops.c src/sys/kern vfs_subr.c ... Message-ID: <200110240811.f9O8BRG94627@apollo.backplane.com> References: <200110230121.f9N1LTa24181@freefall.freebsd.org> <20011024102305.A65744@straylight.oblivion.bg>
next in thread | previous in thread | raw e-mail | index | archive | help
:Would not an MFC of this break binary compatibility for fs-like modules,
:like, badly?
:
:G'luck,
:Peter
I don't know of any binary-only FS kld's but even if there are a few
I don't see any other way of implementing the kern.maxvnodes limits
cleanly. I floated the patchset a while back, which you can examine at:
http://apollo.backplane.com/FreeBSD4/maxvnodes-03.diff
Considering the price of ram we've started to receive reports from
various sources (not just Yahoo) of serious memory-related scaleability
issues. Once you load 2G+ into a machine -stable starts to overload
the vnode cache (mainly when dealing with a large number of small files),
which can result in a machine panic. The cleanest way to fix it is
to enforce kern.maxvnodes and the cleanest way to enforce kern.maxvnodes
is to arrange in the vnodes in an LRU list under the mount structure.
That means changing the existing list from LIST to TAILQ.
We might have to go a step further to solve additional latency issues
with sync - by turning the one vnode list into two lists: a clean and a
dirty list. However, before I explore that possibility I am comitting
the fast-loop fixes discussed in hackers.
-Matt
Matthew Dillon
<dillon@backplane.com>
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200110240811.f9O8BRG94627>