Created attachment 332011 [details] logs from the x86_64 system Description of problem: When I run current nightly installation on RHEL-4, installation ends with 'Could not run gpg.' but exit code is 0. Version-Release number of selected component (if applicable): Satellite-5.3.0-RHEL4-re20090213.1 How reproducible: 2 of 3 attempts (on i386 and x86_64 arch, s390x failed on something else) Steps to Reproduce: 1. # /mnt/redhat/devel/candidate-trees/Satellite-5.3.0-RH EL4-re20090213.1/s390x/s390x//install.pl --answer-file=/mnt/tests/CoreOS/RHN-Sat ellite/Inter-Satellite-Sync/Sanity/general/answers.txt --non-interactive --disco nnected --run-updater Actual results: * Setting up users and groups. ** GPG: Initializing GPG and importing key. ** GPG: Creating /root/.gnupg directory Could not run gpg. Exit value: 2. Please examine /var/log/rhn/rhn-installation.log for more information. Expected results: should work Additional info: # tail /var/log/rhn/rhn-installation.log [...] 5/5 Installing ../EmbeddedDB/oracle-server-admin-0.1-11.el4.noarch.rpm All required packages have been installed gpg: failed to create temporary file `//.gnupg/.#lk0x8545868.dell-pe1850-02.rhts .bos.redhat.com.13697': No such file or directory gpg: //.gnupg: directory created gpg: new configuration file `//.gnupg/gpg.conf' created gpg: WARNING: options in `//.gnupg/gpg.conf' are not yet active during this run gpg: keyring `//.gnupg/pubring.gpg' created # ls /.gnupg gpg.conf pubring.gpg
Err, one more thing: # rpm -q gnupg gnupg-1.2.6-9
I was unsuccessful in recreating this problem. I used my RHEL4u7 x86_64 with a minimal install, then used the same version of the iso as Jan. I also used his answers file (with my email address in it), and it created the gpg information just fine. The one thing I did which I always do, was I registered my machine to webqa and up2date'd it fully BEFORE proceeding with the installation. I'd like to look at what state the machine was in before installation. How do I get access to the rhts machines?
Hello, RHTS machines are installed (and destroyed) on demand, so you can not get access to them (but you can reserve one of them). The only thing we did differently is I did not updated the system before I ran the installer (I have the system registered into the webqa as well). Also I'm running the installer in the environment without TTY? It is possible to simulate it using "cron" or "atd". Could that be the case? I'll test it again, Jan
Hello, I have: # at now at> /mnt/redhat/devel/candidate-trees/Satellite-5.3.0-RHEL4-re20090220.1/i386/i386/install.pl --answer-file=/mnt/tests/CoreOS/RHN-Satellite/Installer/Sanity/install/answers.txt --non-interactive --disconnected --run-updater at> <EOT> job 2 at 2009-02-23 04:55 And that PASSed. I have done exactly what test would do. But when I have scheduled the test again, I got the error again: http://rhts.redhat.com/cgi-bin/rhts/jobs.cgi?id=47640 What could be causing this?
I have no idea yet what causes this, still needs investigation.
Not sure what to do with this, I definitely cannot reproduce under normal install conditions, it must be something to do with the automated nature of the RHTS environment. I notice in the original post: gpg: failed to create temporary file `//.gnupg/.#lk0x8545868.dell-pe1850-02.rhts .bos.redhat.com.13697': No such file or directory gpg: //.gnupg: directory created gpg: new configuration file `//.gnupg/gpg.conf' created gpg: WARNING: options in `//.gnupg/gpg.conf' are not yet active during this run gpg: keyring `//.gnupg/pubring.gpg' created Whereas in a normal install everything is happening in /root/.gnupg What user does the rhts tests get run as? Do they have a home directory on the system? Not real clear how the technology involved here works but hopefully this is of some help.
Ok to bump down the priority/severity of high? I don't think this is an issue customers can hit.
You are right, but this is blocking my RHEL4 testing - I'm planning to test this: http://wiki.test.redhat.com/Faq/TipsAndTricks/FakeTerminal together with: http://cvs.devel.redhat.com/cgi-bin/cvsweb.cgi/tests/RHN-Satellite/Installer/Sanity/install/ As this is the only thing I can think of. Also last few nightlies was not installable on the RHEL4 because of dependency issues. Do you think that can help? So please leave this high/high as it is high/high for a QA.
Heh, just trying your idea: Stated this RHTS job: http://rhts.redhat.com/cgi-bin/rhts/jobs.cgi?id=53380 With these changes: http://cvs.devel.redhat.com/cgi-bin/cvsweb.cgi/tests/RHN-Satellite/rhn-satellite-install.sh.diff?r1=text&tr1=1.51&r2=text&tr2=1.52
You were right! Exporting $USER and $HOME is sufficient to solve this. Interesting is, RHEL5 do not need it. If you want to fix this in the installer somehow or CLOSE this, I'm OK with it. Thanks, Jan
Excellent that is a relief. :) Probably the version of GPG in RHEL 5 using a different mechanism to determine who the current user is. Will close now, cheers.