Date: Wed, 2 Jun 2004 15:22:29 -0700 (PDT) From: Jean-Marc Zucconi <jmz@dalai-zebu.org> To: Christian Hiris <4711@chello.at> Cc: freebsd-questions@freebsd.org Subject: Re: Installworld with NFS mounts Message-ID: <200406022222.i52MMTje096858@freefall.freebsd.org> In-Reply-To: <200406030010.44248.4711@chello.at> References: <200406021741.i52HfQ4k064742@freefall.freebsd.org> <200406030010.44248.4711@chello.at>
next in thread | previous in thread | raw e-mail | index | archive | help
>>>>> Christian Hiris writes: > On Wednesday 02 June 2004 19:41, Jean-Marc Zucconi wrote: >> Is it supposed to work? I have properly mounted all required disks on >> the target machine and it failed first with 'touch: not found'. OK I >> copied touch in /tmp/install.$$ and tried again. Then it failed >> because it could not find install, then rpcgen. I copied them again in >> the temp dir, but now it can't go beyond the following: >> >> ===> lib/csu/i386-elf >> install -o root -g wheel -m 444 crt1.o crti.o crtn.o gcrt1.o /usr/lib >> install: crt1.o: No such file or directory >> >> Any ideas? >> >> Jean-Marc > It works fine for me. I use it for kernel, userland, ports and package > installation. Builds from nfs client side work fine, too. Some points you can > check: > 1. > On your building machine: Does /usr/obj/usr/src/lib/csu/i386-elf/crt1.o exist? Yes of course (excepted that it is in fact /usr/obj/u4/src/lib/csu/i386-elf/crt1.o on both machines because of symlinks) > If not something went wrong with your "make buildworld". No because 'make installworld' worked on the source machine. > On the target machine: Are kernel and the userland you try to install in sync? No but this should not matter since I upgrade from 4.9 to 4.10 (differences are minor) and the context was the same on the source machine. > 2. > From the FreeBSD Handbook: "All the machines in this build set need to > mount /usr/obj and /usr/src from the same machine, and at the same point." > http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/small-lan.html > Are you sure both directory trees exported and mounted correctly? Yes of course. > ie. my /etc/exports line on the nfs server looks like this: > /usr/obj /usr/src /usr/ports -maproot=0:0 -network 192.168.123.0 -mask > 255.255.255.0 > My corresponding /etc/fstab lines on the target host: > 192.168.123.10:/usr/ports /usr/ports nfs rw,-i,-s,-r=8192,-w=8192 0 0 > 192.168.123.10:/usr/src /usr/src nfs rw,-i,-s,-r=8192,-w=8192 0 0 > 192.168.123.10:/usr/obj /usr/obj nfs rw,-i,-s,-r=8192,-w=8192 0 0 > 3. > Does nfs properly work? Some cheap low-end NICs possibly show a high packet > retransmission rate and packet loss. This also can happen, if your nfs server > is overloaded. You can check this with 'netstat -i -w 5', tcpdump or any > other network sniffer. In this case tuning of your nfs mounts could help a > little (man 8 mount_nfs). Try to play around with readsize and timeouts or > try to use tcp instead of udp. This is not a NFS problem. If it was the case I would have noticed it because the target machine also routinely mounts /usr/local and /usr/X11 from the source machine. Jean-Marc -- Jean-Marc Zucconi -- PGP Key: finger jmz@FreeBSD.org [KeyID: 400B38E9]
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200406022222.i52MMTje096858>