Are you sure it works fine Paul? In gmond, the index (key) for stored host data is the host IP address. If your hosts with multiple IP addresses only ever use one of address, then yes, all will be fine.
But if the source IP flips between the IPs occasionally, then any or all of this happens: - The headnode will have entries for each IP seen from the multi NIC host. If you netcat the headnode you will find entries for each IP with each of them having the same hostname. Gmetad does not like this. - As the headnode indexes by IP, each of the entries only gets data some of the time. It will end up looking like these hosts are going up and down. - Unlike gmond, gmetad "indexes" by hostname, so at the very least you will have /var/log/messages filling with the RRD update error of a second data update at the same wallclock time. - At one stage gmetad would also abandon parsing the cluster XML stream at that point, meaning that the cluster view may have many hosts missing. I hacked a fix for this so that gmond saved data indexed by hostname. If you want the hack then a Matt Toy can supply it, but it was for ganglia 3.0.4 regards, richard -- kind regards, Richard Quoting Paul Choi <[EMAIL PROTECTED]>: > I have reverse lookups for both interfaces and Ips. Reverse DNS mapping > is fine. > > Maybe rather than manually overriding hostname, it might be easier if > there's a way to have gmond send packets to other gmond(s) using a > specific interface. > > -Paul > > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of > richard grevis > Sent: Wednesday, September 19, 2007 9:45 PM > To: Matt Cuttler > Cc: [email protected] > Subject: Re: [Ganglia-general] Overriding hostname > > > Are you sure? As I remember it, the hostnames are generated by a reverse > dns lookup on the headnode (or whichever node you poll). You can play by > netcatting (nc) the headnode 8649 as well as the gmetad node 8651 and > having a look. > > I think it is true that a reverse DNS mapping can't be specified in any > /etc/hosts file. Is that correct? At my last job, software did not > generally do reverse lookups, so when ganglia came along it was us that > discovered all the bad reverse DNS entries. > > There have been discussions earlier about getting each gmond to send a > hostname rather that using the source address and reverse DNSing it on > the headnode. > One gets all sorts of boring problems like bad engineering hostnames > rather than nice ones, and if you hasve multiple IP interfaces/NICs on a > node then hostnames may flip around. That makes it look like hosts are > coming up or going down when they are not. > > regards, > Richard > -- > kind regards, > Richard > > > Quoting Matt Cuttler <[EMAIL PROTECTED]>: > > > Paul Choi wrote: > > > Hi, > > > I was wondering if there's a way to override hostnames of the > reported > > > client hosts? > > > For example, > > > I have a host named db02b. (10.1.5.7). And a virtual interface > eth0:0 > > > with 10.1.3.3. The IP 10.1.3.3 maps to another name du02. > > > > > > Right now, Ganglia shows the host as db02b. But I'd like to show it > as > > > du02 (the eth0:0 IP). > > > > If I remember correctly, modifying the /etc/hosts file on the machine > > running gmetad will work.. > > > > -Matt Cuttler > > > > > ------------------------------------------------------------------------ > - > > This SF.net email is sponsored by: Microsoft > > Defy all challenges. Microsoft(R) Visual Studio 2005. > > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > > _______________________________________________ > > Ganglia-general mailing list > > [email protected] > > https://lists.sourceforge.net/lists/listinfo/ganglia-general > > > > ------------------------------------------------------------------------ > - > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2005. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > Ganglia-general mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/ganglia-general > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2005. > http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ > _______________________________________________ > Ganglia-general mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/ganglia-general > ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Ganglia-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ganglia-general
