[BlueOnyx:04659] Re: CMU doesn't restore subdomains

[Michael Stauber]

Hi Maxim Dolgikh,

> I have the same problem when importing sites:
> "
> sh: -c: line 2: syntax error: unexpected end of file
> sh: -c: line 0: syntax error near unexpected token `)'
> sh: -c: line 0: `echo -n "ScanIn: ") as a HASH ref while "strict refs"
> in use at /usr/cmu/perl/Archive/Tar.pm line 624.'
> "
> 
> i also tried restoring one at once, but still have this
> problem. Were you able to solve this? Can you please let me know how.

Yeah, the problem there is as follows:

During a CMU-Import CMU needs to create the site and users first. So it reads 
the associated XML files to gather information about the site name, enabled 
features, usernames and so on. It also reads the disk quota for sites and 
users from the XML files and creates the site accordingly.

Once that's done it imports the tarballs and tries to extract them into the 
respective site and user folders.

Now sometimes (expecially on BlueQuartz - to a much lesser extend on BlueOnyx) 
the quota information on the *old* box (where you created the export) was 
wrong. Or the site (or users of it) were already over quota.

When CMU then tries to extract a 200MB tarball with user data and the user has 
only 20MB disk space allowed, CMU will barf and throw the error you've 
reported.

Work arounds:

Before you create a CMU-Export, make sure that no sites and users are over 
quota.

If that wasn't done, or if it's already too late for that, import all sites 
and users as usual, but note on which sites that failed. Skip them for now. If 
a site was partially created, delete it from the GUI.

Then import only the failed sites with something like this:

cmuImport -c -n "www.site.com" -d /importdirectory

The "-c" switch will import the site and user structure from the XML file, but 
will not extract any tarballs. You then need to manually extract the tarballs 
into the right directories.

Another alternative is to edit the user or site quota in the XML files prior 
to importing.

Long term wise we'll revise CMU a little to prevent all this. We'll try to 
import as "root", which effectively gives the site and users unlimited disk 
quota during the import. Only after the files have been extracted we'll chown 
them back to the right UID and GID. That's how it should have been done to 
begin with <sigh>.

-- 
With best regards

Michael Stauber