Shulou(Shulou.com)06/01 Report--

This article introduces the knowledge of "how to use hashicorp packer to export dbcolinux to virtual machine and docker format". In the operation of actual cases, many people will encounter such a dilemma, so let the editor lead you to learn how to deal with these situations. I hope you can read it carefully and be able to achieve something!

1Perfect packer file, other preparation work

Due to the change of osx plus parallelsdesk, we need to change the basic packer file. The following only refers to the added part, and some parts of article 1 and 2 will be omitted:

In builders, it looks like this:

Type: "parallels-iso", "guest_os_type": "linux", "hard_drive_interface": "ide",. " Parallels_tools_flavor ":" lin ",." Boot_wait ":" 4s "," shutdown_command ":" sudo poweroff ",." Prlctl ": [[" set "," {{.Name}} ","-- startup-view "," window "]]," boot_command ": […" Tce-load-iw parted.tcz, "tce-load-iw grub2.tcz", "tce-load-iw squashfs-tools-4.x.tcz", Integrated here for compiletc,base-dev including linux header "echo 'setup gcccore pkgs...'", "tce-load-iw compiletc.tcz", "echo' setup gccextra pkgs...'", "tce-load-iw perl5.tcz", "tce-load-iw ncurses.tcz", "tce-load-iw ncurses-dev.tcz".]

It can be seen that in order to debug, 1Jing 2 steps to facilitate continuous integration, we also put the entire gcc into bootcommand, but also integrated into some necessary libraries and tools.

Then prepare the files, and packer files in the same layer of the src folder, divided into / base and / others, such a folder arrangement, base corresponds to phase1, is the original source folder used to compile toolchain,64kernel + .x tinycorelinux src repos/busybox-1.19.0patched.tar.gz,busybox-1.19.0-config. And others corresponds to the source code package to be used by phase2. There is autoscan-devices.copenssl-1.0.1.tar.gz,openssh-5.8p1.tar.xz,sudo-1.7.2p6.tar.gz,libzip-0.10.tar.xz,make-3.81.tar.gz,etc...

So, the "provisioners" section would look like this:

{"type": "shell", "pause_before": "1s", "execute_command": "echo'| sudo-S sh-c'{{.Vars}} {{.Path}'", "inline": ["cp-R / tmp/tce ~ /"]}, {"type": "shell", "pause_before": "1s" "execute_command": "echo'| sudo-S sh-c'{{.Vars}} {{.Path}'", "scripts": [". / scripts/1.bootstrap.sh"]}, this tmp needs to be generated in advance because we will no longer release microcore.cpio in our later bootstrap. " Sudo sh-c 'mkdir / mnt/hda1/tmp/' "," sudo sh-c' chown tc:staff / mnt/hda1/tmp/' ". {"type": "file", "source": "src/base", "destination": "/ mnt/hda1/tmp"}, {"type": "shell", "pause_before": "1s", "execute_command": "echo'| sudo-S sh-c'{.Vars}} {{.Path}'", "scripts": [. / scripts/2.1.build64toolchainandkernel.sh ",". / scripts/2.2.build3264rootfsbase.sh "]}}

It can be seen that the original 2script3 3 has been integrated into 2, and the script3 here is to be added. For ease of debugging, it is easy to view it in the output window of packer, including the above packer, which tries to write a simple tree xml in the line. In the following script, we have also made some small changes and adjustments that are different from those in the original article:

We replaced the important notes with echo "ing..." In the script, those who extract the files do not add verbose, which involves the place where toolchain,kernel is compiled. All. / configure CC= "gcc-w", and the above unified > / dev/null, and, toolchain is a separate folder, separate from the basic rootfs required / lib,/lib64, we will also switch native gcc/ newly built cross gcc method to no longer use export CROSS_COMPILE, but directly use the file name.

Debug statements, found that the previous article wrote, a letitdebug meaningless statement, placed at the end of the script or forced breakpoint will not make on error ask break, reboot will produce breakpoints so that the following statements can not run, but will also lose the current environment, packer-debug is still the most available, except that it is a bit annoying to ask every step, packer does not support-debug-on-error.

Most of these need to be changed at the script level, let's do it one by one:

2changes to the script of the bootstrap script

Because we want to form a pure / system, the system under / is removed (some of the comments below), and the whole bootstrap will look like this:

Echo "HD INSTALL..." # below suitable / cd / mnt/hda1/#gunzip-c / mnt/hda1/boot/microcore.gz > / tmp/microcore.cpio#cpio-idmv

< /tmp/microcore.cpio#tce-load不能sudo,只能unsquashfs方式安装#cp -R /tmp/tcloop/squashfs-tools-4.x/usr/ /mnt/hda1/#unsquashfs -f -d /mnt/hda1/ /mnt/hda1/tce/optional/gcc_libs.tcz#unsquashfs -f -d /mnt/hda1/ /mnt/hda1/tce/optional/openssl-0.9.8.tcz#unsquashfs -f -d /mnt/hda1/ /mnt/hda1/tce/optional/openssh.tcz#ldconfig#tar zxf /mnt/hda1/mydata.tgz -C /mnt/hda1/#/下，tce=hda1实测运行tce-load是无效的#echo menuentry \"dbcolinux hd\" { >

> / mnt/hda1/boot/grub/grub.cfg#echo linux / boot/bzImage com1=9600,8n1 loglevel=3 user=tc console=ttyS0 console=tty0 noembed nomodeset root=/dev/hda1 swapfile=hda1 tce=hda1 opt=hda1 home=hda1 norestore > > / mnt/hda1/boot/grub/grub.cfg#echo} > > / mnt/hda1/boot/grub/grub.cfg# is suitable for / system. Mkdir / mnt/hda1/system/# everything except the grub entry needs to be assembled or compiled little by little. Note that root,init is added here, and norestore,copyfiles tce is for restore. There is no need for echo menuentry\ "dbcolinux systemhd\" {> > / mnt/hda1/boot/grub/grub.cfgecho linux / boot/bzImage64 com1=9600,8n1 loglevel=3 user=tc console=ttyS0 console=tty0 noembed nomodeset root=/dev/hda1 init=/system/linuxrc swapfile=hda1 tce=hda1 opt=hda1 home=hda1 norestore > > / mnt/hda1/boot/grub/grub.cfgecho} > > / mnt/hda1/boot/grub/grub.cfg3,patch source code

These files are modified, and in the following scripts will be released to overwrite the corresponding levels in the corresponding source folder, forming the effect of patch. So its structure retains the structure of the original source folder. Then packaged into a xx patch.tgz file, there are: linux-2.6.33.3_patchs,glibc-2.11.1_patchs,busybox-1.19.0_patchs,patched parts, please go to article 3 to compare. You can also go directly to the source code database to find.

4, / scripts/2.build64toolchainandkernel.sh

Here is mainly article 2, the reinforcement and revision of the original script 2jue 3:

…… Echo "compiling binutils." # this is set to shared, because the ld in lib64 needs to highlight that so,ld-* needs to make a copy later. There will be an ar:file truncated problem when compiling 64bit bfd in # shared mode, so it has to be compiled twice. The one in a fix tce-load compiletc, it will be installed in / usr/local/bin/cd.. / binutils-2.20 & & mkdir b & & cd b../configure CC= "gcc-w"-enable-64-bit-bfd > / dev/nullmake > / dev/nullmake install > / dev/nullcd. /.. / binutils-2.20 & & mkdir b2 & cd breadth as long as binutils- enable-shared, then the gcc compiled with it must also be-enable-shared It will affect that the next c libs.cpplibs is all shared../configure CC= "gcc-w"-prefix=/mnt/hda1/system/toolchain-target=$TARGET-enable-shared-disable-multilib-enable-64-bit-bfd > / dev/nullmake > / dev/nullmake install > / dev/null. Echo "compiling c lib startups." # Standard C library header files and some necessary startup files # after all, it's not good to switch cross compile through export cross compile, let's just specify it manually, it's over # if cross compile gcc goes wrong and fenv.h can't find it, it means you've changed some prefix folders, if tls.h can't be found, then you don't use x86_64-pc-linux-gnu-gcc and gcc correctly, compiling c libs Both are x86_64-pc-linux-gnu-gcc rather than gcctar zxf / mnt/hda1/tmp/base/glibc-2.11.1_patchs/Archive.tgz-C / mnt/hda1/tmp/base/glibc-2.11.1/cd. /.. / glibc-2.11.1 & & mkdir b & & cd b # in glibc's eyes, build and host mean the same thing. Different from other standard triples.. / configure CC= "x86_64-pc-linux-gnu-gcc-w"-prefix=/mnt/hda1/system/toolchain/$TARGET-build=$MACHTYPE-host=$HOST-target=$TARGET-with-headers=/mnt/hda1/system/toolchain/$TARGET/include-disable-multilib libc_cv_forced_unwind=yes libc_cv_c_cleanup=yes > / dev/nullmake install-bootstrap-headers=yes install-headers > / dev/nullmake csu/subdir_lib > / dev/nullinstall csu/crt1.o csu / crti.o csu/crtn.o / mnt/hda1/system/toolchain/$TARGET/lib > / dev/null# uses the newly compiled 64gcc to process x86_64-pc-linux-gnu-gcc-nostdlib-nostartfiles-shared-x c / dev/null-o / mnt/hda1/system/toolchain/$TARGET/lib/libc.so > / dev/null... Echo "compiling C++ libs." # Standard C++ Library # this is a bug1ln-s / usr/local/bin/file / usr/bin/filecd. /.. / gcc-4.4.3/b/...... Make CC= "x86_64-pc-linux-gnu-gcc-w" bzImage > / dev/nullmake CC= "x86_64-pc-linux-gnu-gcc-w" modules > / dev/nullmake modules_install INSTALL_MOD_PATH=/mnt/hda1/system > / dev/nullln-s / system/lib/modules/2.6.33.3-tinycore64/kernel / mnt/hda1/system/lib/modules/2.6.33.3-tinycore64/kernel.tclocalcp-f arch/x86/boot/bzImage / mnt/hda1/boot/bzImage645 Scripts/3.build3264rootfsbase.sh

This is the whole new script 3:

Export PATH=$PATH:/usr/local/sbin:/usr/local/bincd / mnt/hda1/tmp/base/tar zxf busybox-1.19.0patched.tar.gzecho "making basic rootfs lib,lib64." # gcc test.c A demo,/usr/bin/ldd to test it, you can find out the dys# on which to run it. Note here Copy 32 ld-* Dylibs part # / mnt/hda1/system/lib/ was created in the previous script to install lib/modules cp / lib/lib*.so* / mnt/hda1/system/lib/cp / lib/ld-2.11.1.so / mnt/hda1/system/lib/ld-2.11.1.sochmod + x / mnt/hda1/system/lib/ld-2.11.1.so# and then set up the mkdir / mnt/hda1/system/lib64/cp / mnt/hda1 under / lib64 / system/toolchain/x86_64-pc-linux-gnu/lib/lib*.so* / mnt/hda1/system/lib64/cp / mnt/hda1/system/toolchain/x86_64-pc-linux-gnu/lib/ld-2.11.1.so / mnt/hda1/system/lib64/ld-2.11.1.sochmod + x / mnt/hda1/system/lib64/ld-2.11.1.socp / mnt/hda1/system/toolchain/x86_64-pc-linux-gnu/lib64 / lib*.so* / mnt/hda1/system/lib64/# link File. / ld-linux.so.2 can be found in ln-s / system/lib/ld-2.11.1.so / mnt/hda1/system/lib/ld-linux.so.2ln-s / system/lib64/ld-2.11.1.so / mnt/hda1/system/lib64/ld-linux-x86-64.so.2echo "compiling busybox." tar zxf / mnt/hda1/tmp/base/busybox-1.19.0_patchs/Archive.tgz-C / Mnt/hda1/tmp/base/busybox-1.19.0/cd busybox-1.19.0make mrproper > / dev/nullcp-f.. / busybox-1.19.0-config configs/i686_defconfigmake i686_defconfig > / dev/null# can also add CROSS_COMPILE=#. Some file encoding problems can lead to the following-become. If / usr/local/bin/ld rpath=/system/lib no such file no such file or directory appears Maybe that's the kind of mistake. It's weird. Make CC= "gcc-Wl,-rpath=/system/lib-Wl. -dynamic-linker=/system/lib/ld-linux.so.2 "CONFIG_PREFIX=/mnt/hda1/system/ install > / dev/nullecho" make basic rootfs data... "mkdir / mnt/hda1/system/devcd / mnt/hda1/system/devmknod console c 5 1mknod null c 1 3mkdir / mnt/hda1/system/etccd / mnt/hda1/system/etctouch fstabecho proc / system/proc proc defaults 00 > > fstabecho sysfs / system/sys sysfs defauts 00 > > fstabtouch inittabchmod + x inittabecho:: sysinit:/system/etc/init. D/rcS > > inittabecho console::respawn:-/system/bin/sh > > inittabecho:: ctrlaltdel:/system/sbin/reboot > > inittabecho:: shutdown:/system/bin/umount-a-r > > inittabmkdir / mnt/hda1/system/etc/init.dcd / mnt/hda1/system/etc/init.dtouch rcSchmod + x rcSecho #! / system/bin/sh > > rcSecho / system/bin/mount-a > > rcSmkdir / mnt/hda1/system/procmkdir / mnt/hda1/system/sys#clean#rm-rf / mnt/hda1/tmp/base/#reboot

-

Finally, when you enter the newly generated grub system entry, you have to export PATH=$PATH:/system/bin:/system/sbin the command line first. It should be that patchs is not done well. Do it next time.

That's all for "how to use hashicorp packer to export dbcolinux to virtual machine and docker format". Thank you for reading. If you want to know more about the industry, you can follow the website, the editor will output more high-quality practical articles for you!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.