I am migrating our Stratum-0 from CentOS 7 to Rocky 8.
The whole content of /srv/cvmfs/ is preserved, since it is on a network filesystem.
I kept also a copy of the entire content of /etc/cvmfs/, and I have restored it.
Now I am trying to recover the content of /cvmfs/ using cvmfs_server import command
But either I am missing something, or I am not using it correctly
I presume you had to not preserve /etc/cvmfs/repositories.d/* … but that looks like a problem with your httpd. Check those access and error logs. You may also be able to get more details by setting CVMFS_SERVER_DEBUG=3.
Just to add some information from private communication with Jose: There has also been an ownership issue as the uid of the user that owned the repository was not used by the new system.
We have seen issues in the past from migrations that left directories with the wrong uid, but it’s not clear yet if that could be also responsible for the error here.
Yes, the problem was with Apache, indeed.
There is a sort of catch 22 situation here. The /etc/httpd/conf.d/cvmfs..conf is supposed to be created by cvmfs_server commands.
But, as I was trying to restore data, only command was “import”, which failed because of that config file was missing.
I ended up creating it manually, and then import worked.
I don’t understand that, because I have never had that problem.
More importantly, although the stratum 0 files are now no longer missing with 404 errors, we’re now seeing this error on stratum 1s:
failed to fetch manifest (8 - bad whitelist)
I can read the file and it is not expired, so that can only mean that the signature does not match the key, which is /etc/cvmfs/keys/egi.eu/egi.eu.pub. So did you make sure that all the repository keys (in this case it’s the .masterkey that matters) were restored from the original stratum 0?