Re: [pcp] install failing on centos 5.8 [SOLVED]

[Ken McDonell <kenj@xxxxxxxxxxxxxxxx>]

On 30/07/13 17:33, Nathan Scott wrote:
> ----- Original Message -----
> > On a brand new system, with t-o-t git repository, Makepkgs then ... kaboom
> > from rpm
> >
> > $ sudo rpm -U `ls build/rpm/*.rpm | sed -e '/src.rpm/d' -e '/sheet2pcp/d'`
> > Rebuilding PMNS ...
> > Starting pmcd ...
> > Starting pmlogger ...
> > Performance Co-Pilot starting inference engine(s) ...
> > Performance Co-Pilot starting pmproxy (logfile is
> > /var/log/pcp/pmproxy/pmproxy.log) ... /etc/init.d/pmproxy: line 198: 17726
> > Segmentation fault      $PMPROXY $OPTS
> > failed (status=139)
> > $ /etc/rc.d/init.d/pmie: pmcd_wait failed ...
> > pmcd_wait: exit status: 2
> >
> > Anyone else seeing anything like this?
> Not here, RHEL5 is looking good for me with tot.  Strangely, line 198 of the
> dev src/pmproxy/rc_pmproxy does not have a $PMPROXY $OPTS invocation - thats
> on line 257.
>
> Could this be a case of configure and/or "make install" getting confused and
> having both /etc/init.d/pmproxy and /etc/rc.d/init.d/pmproxy (with different
> contents)?  And maybe different libexec dir settings with different pmproxy
> binaries too?
That's a reasonable guess ... but I don't think this was the case ... 
pcp.conf looked like the other CentOS 5.x system I have access to 
(except for a PCP_LIB_DIR=/usr/lib64 vs PCP_LIB_DIR=/usr/lib which was 
expected).
Tracked down eventually to a DNS issue ... hostname not in /etc/hosts, 
DNS for hostname -> IP addr that was not the IP addr of any network 
interface on the box.
I'm not going to try and diagnose this further as I have bigger and 
smellier fish to fry.
Apologies for the noise.