sudo docker build -t gmetad-gangliaweb gmetad-gangliaweb/.
Sending build context to Docker daemon 17.41 kB Step 1 : FROM debian:jessie ---> 9dfe4b484ffc Step 2 : MAINTAINER Vladimir Vuksan <[email protected]> ---> Using cache ---> af70ab85d014 Step 3 : COPY config /config ---> Using cache ---> 4d2c6f972b53 Step 4 : RUN apt-get update ---> Using cache ---> 393c08774ebf Step 5 : RUN apt-get upgrade -y ---> Using cache ---> dc16b05415a5 Step 6 : RUN DEBIAN_FRONTEND=noninteractive apt-get install -y apache2 ---> Running in 116a4fe8ca56 Reading package lists... Building dependency tree... Reading state information... E: Unable to locate package apache2 The command '/bin/sh -c DEBIAN_FRONTEND=noninteractive apt-get install -y apache2' returned a non-zero code: 100

Hi:
Sorry for disturbing.
I catch an error using gmetad 3.6,
error message show in /var/log/message is:

  kernel: gmetad[6775]: segfault at 0 ip 000000000040806a sp 00007fadcdc5d2b0 error 4 in gmetad[400000+e000]

Can you set the root html directory to redirect requests to /ganglia.

If my mail server connect must by proxy,what should I add scripts in the ganglia-alert.Thanks.

When I use gmetric.py(v3.1) in gmond 3.1.7 , it does not report any error . But nothing reports to gmond.

/usr/local/bin/gmetric.py --name="t1" --value="10" --type=int32 --group="ttt"

gmetric(v3.1.7) work OK :

./gmetric --name="t1" --value="10" --type=int32

gmond debug:
Processing a metric metadata message from
**Allocating metadata packet for host---- and metric --t1-- ***
saving metadata for metric: t1 host:
Processing a metric value message from
**Allocating value packet for host---- and metric --t1-- ***

Is gmetric.py work OK ?

The original regexp and log format included items that were not supported by recent versions of Tomcat. I changed the log format to be as close to the Tomcat standard string, adding only the information that TomcatLogtailer.py needs.

• Common log format + request duration log format
• More robust regexp that will fail more cleanly

is it possible to pass multiple groups using -g
say, -g group1,group2

gmetric.py --host <host_ip>
--port
--dmax 50
--type float
--name cpu_nice
--units milli seconds
--value 5
--spoof <spoof_ip>:
--group cpu,system <==

it is possible to do this using ganglia plugin

i was trying to have the cli options by running g-l with --help but i got an error instead:

./ganglia-logtailer --help
Traceback (most recent call last):
  File "/yada/yada/ganglia-logtailer", line 50, in <module>
    os.mkdir(logDir)
OSError: [Errno 13] Permission denied: '/var/log/ganglia'

this was with my user, so of course it couldnt create that dir nor access it.

I think g-l shouldnt try to initialize/read logging if only run via --help

simple readme issue:
now - http://blog.vuksan.com/monitoring/2010/09/29/integrating-graphite-with-ganglia/

ganglia-logtailer should prohibit submitting metric names that will break later in the ganglia pipeline. The main ganglia code already does some metric name sanitizing; ganglia-logtailer should apply the same rules to metric names before submitting them.

Not an real issue, but why do you prefer using a ramdisk instead of the /dev/shm filesystem ?
I use a 300+ hosts in my ganglia cluster with many gmetrics (+150k rrd files) and I've done quite the same but using /dev/shm instead of ramdisk :
link /var/lib/ganglia/rrds to /dev/shm/rrds
at the start of gmetad, rsync rrds from disk (/var/lib/ganglia/rrds.sav) to memory (/dev/shm/rrds)
at the stop of gmetad, rsync rrds from /dev/shm/rrds to disk (/var/lib/ganglia/rrds.sav)
crontab to rsync mem to disk every 5mn.

Using /dev/shm could be an issue if you really need memory, but ramdisk sizing is difficult, and it reserve memory even if not used .

my 2 cents...