Closed NickyHeuperman closed 7 years ago
Hi @NickyHeuperman, thanks for rising this issue! My images work fine with around 768mb of memory in a virtual machine. Could you show the packages.both/packages.x86_64? I guess you installed a long list of packages and the generated image is pretty huge?
arch-install-scripts b43-fwcutter btrfs-progs clonezilla crda darkhttpd ddrescue dhclient dialog dmraid dnsmasq dnsutils dosfstools elinks ethtool f2fs-tools fsarchiver gnu-netcat gpm gptfdisk grub hdparm ipw2100-fw ipw2200-fw lftp linux-atm mc mtools nfs-utils nilfs-utils nmap ntfs-3g ntp openssh partclone parted partimage refind-efi rsync smartmontools sudo tcpdump testdisk usb_modeswitch vim-minimal wget pkgfile htop lsof strace ruby iperf3 bash-completion moreutils dfc mdadm ccze screen sysstat ipmitool intel-ucode python-requests python-pyudev python-six python-lxml lshw dmidecode
Just the default packages I think, my current working theory is that ipxe takes up too much ram by comparison. Any idea why nfs loop booting doesn't exactly work? It's hard to debug when it won't even drop me to a console. I guess a workaround could be booting the iso directly or temporarily increasing ram during setup but any insight in the matter would be greatly appreciated. Thank you for your quick reaction. BTW the resulting image is about 566MB give or take
566MB sound okay. In the worst case the system needs twice the size of the sqashfs as ramsize. It gets copied into the ram and afterwards extracted. The ram gets freed, but for a short amount of time it is normal that some memory is allocated. Can you share your ipxe/pxelinux config and NFS share options? maybe there is something we can optimize.
ups wrong button
No worries, shall do as soon as I get access to my setup again, won't be today though, thanks in advanceOn 13 Sep 2017 22:44, Tim Meusel notifications@github.com wrote:ups wrong button
—You are receiving this because you were mentioned.Reply to this email directly, view it on GitHub, or mute the thread.
/etc/exports
/export 10.10.0.0/24(ro,fsid=0,insecure,no_subtree_check,async) /export/nfs 10.10.0.0/24(ro,nohide,insecure,no_subtree_check,async) /export/archrescue 10.10.0.0/24(ro,no_subtree_check)
Where nfs contains installimage and images and archrescue contains the unpacked iso, this is bound to the same folder as /var/www/archrescue pxelinux.cfg
DEFAULT archrescue_nfs LABEL archrescue_http MENU LABEL Arch Linux Live Rescue (HTTP) LINUX http://10.10.0.204/archrescue/arch/boot/x86_64/vmlinuz INITRD http://10.10.0.204/archrescue/arch/boot/x86_64/intel_ucode.img INITRD http://10.10.0.204/archrescue/arch/boot/x86_64/archiso.img APPEND archiso_http_srv=http://10.10.0.204/archrescue/ archisobasedir=arch checksum=y nomodeset systemd.mask=systemd-machine-id-commit systemd.mask=systemd-tmpfiles-setup systemd.mask=ldconfig selinux=0 console=tty0 console=ttyS0,115200 script=/usr/local/bin/start_installimage SYSAPPEND 3
LABEL archrescue_nfs MENU LABEL Arch Linux Live Rescue (NFS) LINUX http://10.10.0.204/archrescue/arch/boot/x86_64/vmlinuz INITRD http://10.10.0.204/archrescue/arch/boot/x86_64/intel_ucode.img INITRD http://10.10.0.204/archrescue/arch/boot/x86_64/archiso.img APPEND archiso_nfs_srv=10.10.0.204:/export/archrescue archisobasedir=arch checksum=y nomodeset systemd.mask=systemd-machine-id-commit systemd.mask=systemd-tmpfiles-setup systemd.mask=ldconfig selinux=0 console=tty0 console=ttyS0,115200 copytoram=n script=/usr/local/bin/start_installimage ip=:::::eth0:dhcp SYSAPPEND 3
As far as I can tell by the documentation of ArchLinux and pxelinux there should be no real difference between the two configs
I have btw tried different ip= parameters, including ip=:::::*:dhcp and ip=:::::ens18:dhcp, archlinux seems to boot, tell me target network is reachable but then hangs on mounting /root/.installimage and /root/images
In customize_airootfs.sh
# add installimage NFS mounts if [ -n "$ISO_NFSSERVER" ]; then echo "${ISO_NFSSERVER}:/export/nfs/install /root/.installimage nfs ro 0 0" >> /etc/fstab echo "${ISO_NFSSERVER}:/export/nfs/images /root/images nfs ro 0 0" >> /etc/fstab mkdir -p /root/{images,.installimage} fi
For the past weekend I've been working on doing a port of LARS to alpine (Knowing alpine uses very little resources and can be quite small on disk) I have succeeded in getting that version of lars working on under 200MB ram. For now I'll close this issue as I'm no longer looking into Arch Linux
I've got an issue/question,
I've been succesfull in building the images and booting from it, I've noticed I need about 1.1GB of ram to succesfully boot over http, arch linux pxe install recommend loop booting over nfs on low memory issues (I'd like to run this on vms with a minimum of 512MB ram). Doing so however won't fully boot the image. It hangs on mounting the nfs shares, something that works fine with the http boot for whatever reason.
My current workflow is to pxe boot as the example given but I was wondering if there is any way I could make this boot over pxe without needing this much ram?