Steve,
Matt told me how to get things working. I set
mcast_channel to a different IP address for
each set of nodes. It works great!
Joe
Steven Wagner wrote:
Joe Griffin wrote:
Hello,
I have two clusters running ganglia/gmetad
wonderfully. Each cluster has it's own name
and gmetad seperates the clusters by those names
(the headnode name).
I have a third cluster which has two types of
nodes within the same cluster (type1 and type2).
But gmetad modifies the rrds in BOTH directories
/usr/local/gmetad/rrds/type1 and type2.
I see nodes1-4_type1 and nodes1-4_type2 in
BOTH /usr/local/gmetad/rrds/type1 and type2.
So gmetad displays information like I have twice as many nodes.
The gmetad files are the same as on my two seperate clusters
with headnodes. The only difference is that
ALL of these nodes are on the same cluster. I
just wanted to allow groupings within a cluster.
"CLUSTER NAME" is getting set right from gmond:
Headnode:~# telnet node1_type1 8649 | grep 'CLUSTER NAME'
<!ATTLIST CLUSTER NAME CDATA #REQUIRED
<CLUSTER NAME="type1" LOCALTIME="1026510618">
Headnode:~# telnet node1_type2 8649 | grep 'CLUSTER NAME'
<!ATTLIST CLUSTER NAME CDATA #REQUIRED
<CLUSTER NAME="type2" LOCALTIME="1026510618">
Can I have different groups if the nodes are on
the same cluster (ie the first 3 parts of the IP address are the same)?
Any thoughts would be appreciated,
Thanks,
Joe
I think the current implementation of gmetad only allows for one "layer"
of grouping (i.e. the Metacluster -> individual clusters -> individual
hosts). Although there is no real reason you couldn't hack up gmetad to
check a second config file to determine what "group" a host was in...
the other solutions I can think of all involve duplicating host data
collection.
Alternately, you could run gmetric as part of each gmond startup script
(or a gexec/cron job) to set a local metric named "group_num" and then
hack gmetad to sort by this field.
Or hack the front-end to understand grouping.
Have we just gotten a volunteer to join the gmetad/frontend development
team? :)
