[Spacewalk-list] Monitoring broken in 1.4?

[Jan Pazdziora]

On Thu, May 12, 2011 at 03:56:34PM +0100, David Nutter wrote:
> Hi,
> 
> I'm just wondering if monitoring in 1.4 is known to be broken? I can't
> find anything in the release notes to that effect but I may be blind.
> 
> Everything was fine until I upgraded to 1.4 (from 1.2 via 1.3) and now
> after a morning of fiddling with various perl modules I can get a
> config push to work but none of my probes update. There's also no
> connections from spacewalk to rhnmd on my various hosts so it's not
> surprising that these probes fail to update. 
> 
> I can't find where the monitoring subsytem makes a connection to RHNMD
> on the remote host so I can't proceed further with debugging. Where
> should I look. I can't see anything in the numerous logs in
> /var/log/nocpulse/* but I guess I may need to up logging levels
> somewhere. 
> 
> Background detail:
> 
> Monitoring broke for me with 1.4, initially because a scout config
> push would fail with:
> 
> 18:25:07 2011: ERROR: Spacewalk 1: Cannot generate configuration:
>    Message SatClusterRecord::get_physical_location_name() not
>    understood
> 
>    TRACEBACK: at
>      /usr/lib/perl5/vendor_perl/5.8.8/NOCpulse/Object.pm line 443
>         NOCpulse::Object::doesNotUnderstand('SatClusterRecord=HASH(0x2b35fe1809b0)', 'SatClusterRecord', 'get_physical_location_name') called at
> 	    /usr/lib/perl5/vendor_perl/5.8.8/NOCpulse/Object.pm line 463\n\tNOCpulse::Object::AUTOLOAD('SatClusterRecord=HASH(0x2b35fe1809b0)') called at
> 	        /usr/lib/perl5/vendor_perl/5.8.8/NOCpulse/SatConfig/GenerateConfig.pm line 44
> 		   eval {...} called at /usr/lib/perl5/vendor_perl/5.8.8/NOCpulse/SatConfig/GenerateConfig.pm line 19 
>                     NOCpulse::SatConfig::GenerateConfig::handler('Apache2::RequestRec=SCALAR(0x2b3606ce57d0)') called at -e line 0\n\teval {...} called at -e line 0
> 
> I had to fix a few things in GenerateConfig.PM and NPRecords.pm; plus
> a method invocation in RHN/DB.pm. I'd send a patch but some (most?) of
> this has already been done upstream. Changed lines on my install are:
> 
> /usr/lib/perl5/vendor_perl/5.8.8/NOCpulse/SatConfig/GenerateConfig.pm:43:        my $location = $satCluster->get_PHYSICAL_LOCATION_NAME; ##DN CHANGED
> /usr/lib/perl5/vendor_perl/5.8.8/NOCpulse/SatConfig/GenerateConfig.pm:109:#    NOCpulse::DBRecord->Disconnect(); ##DN Changed
> /usr/lib/perl5/vendor_perl/5.8.8/NOCpulse/SatConfig/GenerateConfig.pm:155:    #NOCpulse::DBRecord->Disconnect(); ##DN Changed
> /usr/lib/perl5/vendor_perl/5.8.8/NOCpulse/NPRecords.pm:338:                                          $self->commandLine($self->get_HOSTADDRESS, ##DN Changed
> /usr/lib/perl5/vendor_perl/5.8.8/RHN/DB.pm:373:    $dbh->set_err(9999,"Can't disconnect cache-loaded RHN::DB handle"); #DN Changed

The fix in Spacewalk master is commit

	dd69d8fe96a1337a65d8d645f9ecd0d9b05342d0.

> Then the scheduleEvents script would run without error. After that the
> config push would still fail with errors like:
> 
> Cannot POST to http://spacewalk.bioss.sari.ac.uk/cgi-bin/upload_results.cgi:
> 
> This appeared to be because the webserver was looking in
> /var/www/cgi-bin/ rather than /usr/share/nocpulse/cgi-bin/ for the
> scripts fetch_nocpulseini*.cgi, upload_file.cgi and
> upload_results.cgi. 
> 
> So, I symlinked /var/www/cgi-bin to /usr/share/nocpulse/cgi-bin/. Push

I assume you have edited some Apache .conf files and now have some
.rpmnew files around, namely rhn_monitoring.conf -- we've changed
the Alias target to point under /usr/share/nocpulse.

-- 
Jan Pazdziora
Principal Software Engineer, Satellite Engineering, Red Hat