Bernard, Thanks, but to clarify: I basically did that, it was unzipped,
mounted as a loop device, modified the function script and then I recreated initrd.img with mkcramfs then gzipped it. Are you saying that I should not create a cramfs file from the initrd
directory? Otherwise the steps in
your response are the same as performed. Thanks, Bela -----Original Message----- Hi Bela: It appears that you can
mount the initrd.img and write to it: mv initrd.img
initrd.img.gz gunzip initrd.img.gz mount -o loop initrd.img
/mnt Modify the file, then
gzip it. Hopefully it'll work. Cheers, Bernard From: [EMAIL PROTECTED]
on behalf of B. Erdelyi Bernard, My thinking is that for
whatever reason this has to do with BusyBox and the
environment that it is running on our machine. The options are not being
passed on. To test this theory I wanted to hardwire the mount tmpfs command in the script to see what happens. I
found the command in the “functions” script and replaced it with: mount tmpfs
/new_root –t tmpfs
–o size=2g,nr_inodes=2000,mode=777 || shellout Then I recreate the initrd.img with mkcramfs, gziped it and placed it in the standard flavor
directory. Unfortunately, when I load from the new autoinstall
cd I get an error indicating that the file system is
read only and the rcS script cannot write a text file
to it (line 41 of the rcS script). What am I doing
wrong? I thought that the mode command with 777 would create the file
system with full access. Is there other options being passed on via the $tmpfs_opt variable? Or should I be setting rw access somewhere else when I recreate the cramfs initrd.img? I would really like to test
this out to see if by bypassing the parameter passing solves the issue.
It may give you a better hint at what is going on. Any suggestions will
be appreciated. Thanks, Bela -----Original Message----- Hi Bela: If the
tmpfs parameters were indeed passed onto the kernel, then the tmpfs size should
be what you specified. I am really at a loss as to what has gone
wrong... Cheers, Bernard From: [EMAIL PROTECTED]
on behalf of B. Erdelyi Bernard, I
finally installed and tested the new build. One problem has gone away
which is the setting of append options. By specifying the tmpfs size and inodes in the
–append string they now get passed on to the syslinux.cfg
file without having to modify the script. However,
the autoinstall still fails at the target with the
same 99% full message. The tmpfs is created at
the same size as before: 452788. The tmpfs size
specification has no effect on this. I did
not reinstall the client side of systemimager, only
the server side and used the same golden master image file as previously
created. I don’t believe such should have any impact on this issue
since the autoinstall cd is
created on the server. I believe the problem lies with the initial linux kernel in memory. Let me know if I am missing
anything and whether I should reinstall the client side and recreate the golden
image. Thanks, Bela -----Original Message----- Hi Bela: The RPMs
have the same names, but they are built at different times (with slightly
different code). This is my bad so from now on I will give them unique
identifiers such that we can easily identify which version you are using. I guess
the only difference between your installation and my installation is that I do
NOT run si_prepareclient etc. to create a golden image, but instead I use
systeminstaller to create the image. Also, I am using the stock
SystemImager kernel/initrd.img (provided by the i386boot-standard package) but
for your case, you are probably using RHEL3.1's kernel + an autogenerated
initrd.img (via si_prepareclient/UseYourOwnKernel). I am not
sure how much of a difference this is, but I will test this after hearing your
results on Tuesday. Cheers, Bernard |
- RE: [Sisuite-users] Problem with mkautoinstallcd tmpfs Bernard Li
- RE: [Sisuite-users] Problem with mkautoinstallcd tmpfs Bernard Li
- RE: [Sisuite-users] Problem with mkautoinstallcd tmpfs B. Erdelyi
- RE: [Sisuite-users] Problem with mkautoinstallcd tmpfs Bernard Li
- RE: [Sisuite-users] Problem with mkautoinstallcd tmpfs Bernard Li
- RE: [Sisuite-users] Problem with mkautoinstallcd tmpfs Bernard Li
- RE: [Sisuite-users] Problem with mkautoinstallcd tmpfs B. Erdelyi
- RE: [Sisuite-users] Problem with mkautoinstallcd tmpfs Bernard Li