I've been trying to get 3.0.2 installed and working and after spinning
my head far too many times, I've decided I don't really understand the
larger picture of how ganglia works.
Seems there are three pieces to this:
* gmond - daemon that runs on each host, collects information on localhost
* gmetad - "Federation in Ganglia is achieved using a tree of
point-to-point connections amongst representative cluster nodes to
aggregate the state of multiple clusters." Ah - collects data from all
the other nodes? Can it be that important, since by default it is not
even built?
* web app - Some PHP code that sends a request to a host/socket and
displays the output. Implies to me that the target software must have
access to some place where all the data is saved.
What's going on here?
gmond collects stats for localhost. Are these saved somewhere?
Apparently not, since I see no reference to a path. Are they sent
somewhere? Well, gmond.conf provides udp_send_channel, so that suggests
maybe the data is sent elsewhere via this.You can apparently specify a
specific host/port or broadcast it around. This seems reasonable.
But gmond.conf provides udp_recv_channel too. What the heck can this be
for? Just as confusing to me is tcp_accept_channel. So maybe gmond CAN
save data. Now I'm back to my original confusion.
Just as confusing to me is the collection of ports (I may have these
wrong, I've been screwing around with this so much, I may have lost the
defaults)
webapp - 8652 (how to ask for all the collected data ?)
gmond - 8649 (send, receive, accept - only send makes sense to me)
gmetad - 8655,8651 are mentioned near data_source
gmetad - 8651 (answer requests for XML) What XML? from who, for what?
gmetad - 8652 (answer queries for XML) Same questions.
I've read everything I can find on the web site, but I cannot get my
poor head around what's going on so I can figure out why Ganglia doesn't
work for me. Apparently the mail archives are broken so I've not been
able to search the past Emails.
One issue I'm sure must be a factor is that my cluster machines are on
an internal private network and the web access for the summary must come
from outside that network. That's why I've been focusing on how the data
is saved.
If anyone can provide a clearer picture to me of what the data flow is,
I'd appreciate it. TIA
--
=============================================================
Terry Gliedt [EMAIL PROTECTED] http://www.hps.com/~tpg/
Biostatistics, Univ of Michigan Personal Email: [EMAIL PROTECTED]
