Linux-NFS Archive on lore.kernel.org
 help / color / Atom feed
From: bfields@fieldses.org (J. Bruce Fields)
To: "J. Bruce Fields" <bfields@redhat.com>
Cc: NeilBrown <neilb@suse.de>, linux-nfs@vger.kernel.org
Subject: Re: [PATCH 2/2] nfsd: give out fewer session slots as limit approaches
Date: Thu, 19 Sep 2019 14:41:39 -0400
Message-ID: <20190919184139.GG26654@fieldses.org> (raw)
In-Reply-To: <20190919171730.GB333@pick.fieldses.org>

On Thu, Sep 19, 2019 at 01:17:30PM -0400, J. Bruce Fields wrote:
> 	- Maybe we should just keep allowing small sessions (1 slot?)
> 	  even past the limit.  Worst case the subsequent kmalloc fails.

So to be clear, I think that's what I'd do for now instead of trying to
find a better error return.

And probably put the 1/3->1/16 change in a separate patch.

--b.

> 
> --b.
> 
> > 
> > >  Also, I'd like to suggest that the '1/3' heuristic be change to 1/16.
> > >  Assuming 30 slots get handed out normally (which my testing shows -
> > >  about 2k each, with an upper limit of 64k):
> > >    When 90 slots left, we hand out
> > >     30 (now 60 left)
> > >     20 (now 40 left)
> > >     13 (now 27 left)
> > >      9 (now 18 left)
> > >      6 (now 12 left)
> > >      4 (now 8 left)
> > >      2 (now 6 left)
> > >      2 (now 4 left)
> > >      1
> > >      1
> > >      1
> > >      1
> > >  which is a rapid decline as clients are added.
> > >  With 16, we hand out 30 at a time until 480 slots are left (30Meg)
> > >  then: 30 28 26 24 23 21 20 19 18 6 15 15 14 13 12 11 10 10 9 9 8 8 7 7
> > >     6 6 5 5 5 5 4 4 4 3 3 3 3 3 3 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1
> > >     1 1
> > >  slots per session
> > > 
> > >  Am I convincing?
> > > 
> > > To make it more concrete: this is what I'm thinking of.  Which bits do
> > > you like?
> > 
> > Except for the error return, it looks good to me.
> > 
> > --b.
> > 
> > > 
> > > Thanks,
> > > NeilBrown
> > > 
> > > diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> > > index 7857942c5ca6..5d11ceaee998 100644
> > > --- a/fs/nfsd/nfs4state.c
> > > +++ b/fs/nfsd/nfs4state.c
> > > @@ -1573,11 +1573,15 @@ static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
> > >  	total_avail = nfsd_drc_max_mem - nfsd_drc_mem_used;
> > >  	avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION, total_avail);
> > >  	/*
> > > -	 * Never use more than a third of the remaining memory,
> > > +	 * Never use more than a 1/16 of the remaining memory,
> > >  	 * unless it's the only way to give this client a slot:
> > >  	 */
> > > -	avail = clamp_t(unsigned long, avail, slotsize, total_avail/3);
> > > +	avail = clamp_t(unsigned long, avail, slotsize, total_avail/16);
> > >  	num = min_t(int, num, avail / slotsize);
> > > +	if (nfsd_drc_mem_used + num * slotsize > nfsd_drc_max_mem)
> > > +		/* Completely out of space - sorry */
> > > +		num = 0;
> > > +
> > >  	nfsd_drc_mem_used += num * slotsize;
> > >  	spin_unlock(&nfsd_drc_lock);
> > >  
> > > @@ -3172,7 +3176,7 @@ static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfs
> > >  	 */
> > >  	ca->maxreqs = nfsd4_get_drc_mem(ca);
> > >  	if (!ca->maxreqs)
> > > -		return nfserr_jukebox;
> > > +		return nfserr_resource;
> > >  
> > >  	return nfs_ok;
> > >  }
> > 
> > 

  reply index

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-09-25 13:21 [PATCH 0/2] tweak knfsd session slot table sizing J. Bruce Fields
2017-09-25 13:21 ` [PATCH 1/2] nfsd: increase DRC cache limit J. Bruce Fields
2017-09-25 13:21 ` [PATCH 2/2] nfsd: give out fewer session slots as limit approaches J. Bruce Fields
2019-09-19  1:08   ` NeilBrown
2019-09-19 16:22     ` J. Bruce Fields
2019-09-19 17:17       ` J. Bruce Fields
2019-09-19 18:41         ` bfields [this message]
2019-09-20  6:15           ` [PATCH 1/2] nfsd: handle drc over-allocation gracefully NeilBrown
2019-09-20  6:33             ` [PATCH 1/2 - vers2] " NeilBrown
2019-09-20  6:36               ` [PATCH - 2/2] nfsd: degraded slot-count more gracefully as allocation nears exhaustion NeilBrown
2019-09-20 16:28               ` [PATCH 1/2 - vers2] nfsd: handle drc over-allocation gracefully J. Bruce Fields

Reply instructions:

You may reply publically to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190919184139.GG26654@fieldses.org \
    --to=bfields@fieldses.org \
    --cc=bfields@redhat.com \
    --cc=linux-nfs@vger.kernel.org \
    --cc=neilb@suse.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-NFS Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-nfs/0 linux-nfs/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-nfs linux-nfs/ https://lore.kernel.org/linux-nfs \
		linux-nfs@vger.kernel.org
	public-inbox-index linux-nfs

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-nfs


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git