This short tutorial explains the basic parts of print_host_and_vm_rrd.py, an example script demonstrating how to download, parse, and use the operational metrics that are stored in a RRD database on a XenServer host.
The script print_host_and_vm_rrd.py uses the XenServer Python module XenAPI.py for querying the HTTP(S) interface of the XenServer host, and the example python script parse_rrd.py for parsing the xport XML that is returned from the query.
First we need to import both the XenAPI and the parse_rrd.py libraries into the python script.
#!/usr/bin/env/python
import XenAPI
import parse_rrdIn order to download RRD updates, we need to create a XenAPI Session and pass the Session opaque reference in the request as a query parameter. To authenticate against the target XenSever host we use the XenAPI's call login_with_password.
def main():
url = "https://<server>"
session = XenAPI.Session(url)
session.xenapi.login_with_password('root','<password>')Then we create a RRD object and fill it with data. In the following snippet, the params array is overriding the default parameters specified in parse_rrd.py. Note that, if no url is specified, parse_rrd.py uses https://localhost by default.
rrd_updates = parse_rrd.RRDUpdates()
params = {}
params['cf'] = "AVERAGE"
params['start'] = int(time.time()) - 10 #This is for the purposes of this tutorial
params['interval'] = 5
params['host'] = ""
rrd_updates.refresh(session.handle, params, url)It is then very easy to make use of the calls provided in parse_rrd.py to extract and search through the downloaded data. You can see from the code below that for getting VM data we need only specify the VM's uuid, the metric we want to get, and the row we're interested in. This method loops through all the metrics for a given VM and returns only the most recent value for each metric.
def print_latest_vm_data(rrd_updates, uuid):
print "**********************************************************"
print "Got values for VM: "+ uuid
print "**********************************************************"
for param in rrd_updates.get_vm_param_list(uuid):
if param != "":
max_time = 0
data=""
for row in range(rrd_updates.get_nrows()):
epoch = rrd_updates.get_row_time(row)
dv = str(rrd_updates.get_vm_data(uuid,param,row))
if epoch > max_time:
max_time = epoch
data = dv
nt = time.strftime("%H:%M:%S", time.localtime(max_time))
print "%-30s (%s , %s)" % (param, nt, data)The script print_host_and_vm_rrd.py contains also a similar method for printing host data. Note that, in order to download RRD updates for the host, we need to specify it in the params array:
params['host'] = 'true'The result of running this method on every VM present on the target host will produce something similar to this:
**********************************************************
Got values for VM: cafd67c3-ede7-4257-99bb-84d5b6756bb1
**********************************************************
vbd_xvdb_write (12:15:00 , 0.0)
memory_target (12:15:00 , 412090368.0)
vbd_xvdb_read (12:15:00 , 0.0)
cpu6 (12:15:00 , 0.0)
cpu7 (12:15:00 , 0.0)
cpu4 (12:15:00 , 0.0)
cpu5 (12:15:00 , 0.0)
cpu2 (12:15:00 , 0.0089)
cpu3 (12:15:00 , 0.002)
cpu0 (12:15:00 , 0.0104)
cpu1 (12:15:00 , 0.0034)
vbd_xvda_write (12:15:00 , 0.0)
vbd_xvda_read (12:15:00 , 0.0)
memory (12:15:00 , 412090368.0)
**********************************************************
Got values for VM: a9a488ec-2cb4-3ff3-77be-8d8cc453e32f
**********************************************************
vbd_xvdb_write (12:15:00 , 0.0)
memory_target (12:15:00 , 268435456.0)
vbd_xvdb_read (12:15:00 , 0.0)
memory_internal_free (12:15:00 , 113700.0)
memory (12:15:00 , 268435456.0)
vbd_xvda_write (12:15:00 , 808.2336)
cpu0 (12:15:00 , 0.0014)
vif_0_tx (12:15:00 , 0.0)
vbd_xvda_read (12:15:00 , 0.0)
vif_1_tx (12:15:00 , 0.0)
vif_0_rx (12:15:00 , 475.4616)
vif_1_rx (12:15:00 , 0.0)
An easy way to plot the RRD data is by using the Gnuplot utility.
Create a method that outputs the RRD metric data in a format acceptable to Gnuplot. (We are printing out a single data point per row, separating the time and value columns with whitespace.)
def build_vm_graph_data(rrd_updates, vm_uuid, param):
time_now = int(time.time())
for param_name in rrd_updates.get_vm_param_list(vm_uuid):
if param_name == param:
data = "#%s Seconds Ago" % param
for row in range(rrd_updates.get_nrows()):
epoch = rrd_updates.get_row_time(row)
data = str(rrd_updates.get_vm_data(vm_uuid, param_name, row))
data += "\n%-14s %s" % (data, time_now - epoch)
return dataCall this method and save the result to a file. In this example we will be plotting the metrics for 'cpu0'.
for uuid in rrd_updates.get_vm_list():
param = 'cpu0'
data = build_vm_graph_data(rrd_updates, uuid, param)
fh = open("%s-%s.dat" % (uuid, param), 'w')
fh.write(data)
fh.close()Now use Gnuplot to import this data file and plot a graph
gnuplotNote: You can supply the following as an argument to Gnuplot which would help you to automate the process of creating graphs in an easier way
set xlabel "Seconds Ago"
set ylabel "CPU Utilisation"
plot <DATAFILE.dat> using 0:1 title 'cpu0' with linespointsThis should then produce a plot similar this one:
