Date: Sat, 19 Feb 2011 08:56:45 -0800 (PST) From: Kirill Yelizarov <ykirill@yahoo.com> To: freebsd-stable@freebsd.org Subject: Re: NFS client over udp Message-ID: <795118.6346.qm@web120517.mail.ne1.yahoo.com> In-Reply-To: <600887.27323.qm@web120520.mail.ne1.yahoo.com>
next in thread | previous in thread | raw e-mail | index | archive | help
=0A=0A--- On Fri, 2/18/11, Kirill Yelizarov <ykirill@yahoo.com> wrote:=0A= =0A> From: Kirill Yelizarov <ykirill@yahoo.com>=0A> Subject: Re: NFS client= over udp=0A> To: freebsd-stable@freebsd.org=0A> Date: Friday, February 18,= 2011, 5:09 PM=0A> =0A> =0A> --- On Fri, 2/18/11, Jeremy Chadwick <freebsd@= jdc.parodius.com>=0A> wrote:=0A> =0A> > From: Jeremy Chadwick <freebsd@jdc.= parodius.com>=0A> > Subject: Re: NFS client over udp=0A> > To: "Kirill Yeli= zarov" <ykirill@yahoo.com>=0A> > Cc: freebsd-stable@freebsd.org=0A> > Date:= Friday, February 18, 2011, 5:05 PM=0A> > On Fri, Feb 18, 2011 at 05:27:00A= M=0A> > -0800, Kirill Yelizarov wrote:=0A> > > I have a reproducible memory= leak when using nfs=0A> > client with an old=0A> > > nfs server using udp = protocol. I'm running rsync=0A> every=0A> > hour from a=0A> > > mounted nfs= volume to an ufs local volume. Each=0A> time=0A> > rsync is running=0A> > = > wired memory is increased. Wired memory is rising=0A> when=0A> > rsync is= in=0A> > > getblk or biord state. When rsync is complete=0A> memory=0A> > = is not freed and=0A> > > lately all memory is in active or wired state=0A> = and=0A> > system starts=0A> > > swapping. I didn't see such behavior when c= lient=0A> os=0A> > was RELEASE-8.0.=0A> > > Then i switched to STABLE-8.1 o= f late summer and=0A> found=0A> > this leak so i=0A> > > upgraded to STABLE= -8.2Prerelease and got the=0A> same=0A> > leak. vmstat -m=0A> > > shows not= hing big. =0A> > > =0A> > > client version: FreeBSD imap1.***.com=0A> 8.2-P= RERELEASE=0A> > FreeBSD=0A> > > 8.2-PRERELEASE #0: Thu Feb 17 13:54:25 MSK = 2011=0A> > > root@imap1.****.com:/usr/obj/usr/src/sys/IMAP1=A0=0A> > amd64= =0A> > > =0A> > > i tried different mounting options:=A0 =A0=0A> > =A0=A0= =A01 -=0A> > >=0A> >=0A> rw,rsize=3D32768,wsize=3D32768,nfsv3,intr,mntudp,r= dirplus,=0A> > > readdirsize=3D65536,noauto,noexec 2 -=0A> > rw,nfsv3,mntud= p,noexec=0A> > > =0A> > > i also tried setting -tso -txcsum -rxcsum for=0A>= igb=0A> > network card=0A> > =0A> > Is ZFS in use on the system which sees= rising wired=0A> > memory?=0A> No, ufs only. =0AI found an old post statin= g there is a leak with nfs udp client over zfs:=0Ahttp://lists.freebsd.org/= pipermail/freebsd-fs/2010-February/007876.html=0AIn my case i don't have zf= s on server or client but the mount is over udp.=0AHere is top after a day = of hourly rsync=0Alast pid: 17306; load averages: 0.00, 0.00, 0.00 u= p 0+22:07:34 19:52:53=0A53 processes: 1 running, 52 sleeping=0ACPU: %= user, % nice, % system, % interrupt, % idle=0AMem: 12G Act= ive, 1306M Inact, 1609M Wired, 748M Cache, 1641M Buf, 55M Free=0ASwap: 16G = Total, 14M Used, 16G Free=0A=0Aand mbufs used=0A8193/1722/9915 mbufs in use= (current/cache/total)=0A8192/1264/9456/25600 mbuf clusters in use (current= /cache/total/max)=0A8192/605 mbuf+clusters out of packet secondary zone in = use (current/cache)=0A0/768/768/12800 4k (page size) jumbo clusters in use = (current/cache/total/max)=0A0/0/0/6400 9k jumbo clusters in use (current/ca= che/total/max)=0A0/0/0/3200 16k jumbo clusters in use (current/cache/total/= max)=0A18432K/6030K/24462K bytes allocated to network (current/cache/total)= =0A0/0/0 requests for mbufs denied (mbufs/clusters/mbuf+clusters)=0A0/0/0 r= equests for jumbo clusters denied (4k/9k/16k)=0A0/0/0 sfbufs in use (curren= t/peak/max)=0A0 requests for sfbufs denied=0A0 requests for sfbufs delayed= =0A0 requests for I/O initiated by sendfile=0A0 calls to protocol drain rou= tines=0A=0AKirill=0A=0A=0A
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?795118.6346.qm>