Date: Wed, 5 Jan 2022 21:04:14 +0000 From: Rick Macklem <rmacklem@uoguelph.ca> To: Zaphod Beeblebrox <zbeeble@gmail.com>, freebsd-current <freebsd-current@freebsd.org> Subject: Re: Writing large build logs to NFS extremely slow? Message-ID: <YQXPR0101MB0968EAB78E1C05788035D38BDD4B9@YQXPR0101MB0968.CANPRD01.PROD.OUTLOOK.COM> In-Reply-To: <CACpH0MeQe4SYeV-qEYjvvv%2BPbQfNbrNBW=9f4yshn2_v6adGjg@mail.gmail.com> References: <20211007021643.bwglyvrswk2nm3fl@nexus.home.palmen-it.de> <CACpH0MeQe4SYeV-qEYjvvv%2BPbQfNbrNBW=9f4yshn2_v6adGjg@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Zaphod Beeblebrox wrote:=0A= > Is the NFS mounted filesystem NFS?=0A= > I've found NFS mounted ZFS has several=0A= > pathologies like this when there is no SSD cache and/or log vdevs attache= d.=0A= Just fyi, I've recently committed a change to main that changes the NFS=0A= client so that it does append writes directly to the NFS server.=0A= This improved the append write performance by 10x for UFS, but only=0A= improved 30% for ZFS without any ZIL log and sync=3Denabled.=0A= Append writes over NFS is still MUCH slower than a local file system.=0A= (Partially because it has to do every write to the serve FileSync, which me= ans=0A= the server must commit the changes to stable storage before replying to th= e=0A= RPC. It also must do a Getattr RPC against the server before every write,= =0A= to get the up-to-date file size.)=0A= =0A= This change won't by MFC'd for a couple of months and is somewhat broken=0A= in main at this time. I am working with kib@ to fix it.=0A= =0A= rick=0A= =0A= On Wed, Oct 6, 2021 at 10:18 PM Felix Palmen <felix@palmen-it.de> wrote:=0A= =0A= > Hi all,=0A= >=0A= > I use a -CURRENT bhyve vm for testing port builds with poudriere. As=0A= > this vm is only running when needed, but I want to always have access to= =0A= > the build logs, I use NFS to mount /usr/local/poudriere/data/logs from=0A= > the host.=0A= >=0A= > I noticed some few ports take ridiculously long to build while barely=0A= > using any CPU time at all. On a closer look, that's all ports producing= =0A= > a lot of compiler (warning) output, e.g. gcc, gnutls, gtk2, =85=0A= >=0A= > So I assume appending to a large file via NFS gets slower and slower. Is= =0A= > there any mount option I could try to fix this? Right now I only have=0A= > `nolockd`, I also tried `noncontigwr` which didn't change anything.=0A= >=0A= > Thinking about alternatives to NFS, are there any news for client-side=0A= > 9p virtfs? I found <https://github.com/swills/virtfs-9p-kmod>; which=0A= > still builds with a few minor adaptions, but trying to mount a 9p share= =0A= > freezes the machine.=0A= >=0A= > Would you suggest a different mailing list to ask?=0A= >=0A= > BR, Felix=0A= >=0A= > --=0A= > Dipl.-Inform. Felix Palmen <felix@palmen-it.de> ,.//..........=0A= > {web} http://palmen-it.de {jabber} [see email] ,//palmen-it.de=0A= > {pgp public key} http://palmen-it.de/pub.txt // """""""""""=0A= > {pgp fingerprint} A891 3D55 5F2E 3A74 3965 B997 3EF2 8B0A BC02 DA2A=0A= >=0A=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?YQXPR0101MB0968EAB78E1C05788035D38BDD4B9>