1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
|
== INSTALL ==
A.1 Linux Installation (rpm, deb)
.2 Mac OS X Installation
.3 AIX Installation
.4 Solaris Installation
.5 Windows Installation
B. Building from source
C. Post-install steps
D. Non-default build, install and run
This document describes how to configure and build the open source PCP
("pcp") package from source, and how to install and finally run it.
== A.1 Linux Installation
If you are using Debian, or a Debian-based distribution like Ubuntu,
PCP is included in the distribution (as of late 2008). Run:
# apt-get install pcp
If you are using a RPM based distribution and have the binary rpm:
# rpm -Uvh pcp-*.rpm
... and skip to the final section (below) - "Post-install steps".
Special note for Ubuntu 8.04, 8.10, 9.04, 9.10 and 10.04
I've had to make the changes below to /usr/bin/dpkg-buildpackage.
Without these two changes, my pcp builds produce bad binaries with a
bizarre array of failure modes!
#kenj# my $default_flags = defined $build_opts->{noopt} ? "-g -O0" : "-g
-O2";
my $default_flags = defined $build_opts->{noopt} ? "-g -O0" : "-g -O0";
my %flags = ( CPPFLAGS => '',
CFLAGS => $default_flags,
CXXFLAGS => $default_flags,
FFLAGS => $default_flags,
#kenj# LDFLAGS => '-Wl,-Bsymbolic-functions',
LDFLAGS => '',
);
Without these changes, we see QA failures for 039, 061, 072, 091, 135,
147 and 151 ... and the QA 166 goes into a loop until it fills up the
root filesystem.
-- Ken
== A.2 Mac OS X Installation
The only unusual step in installing PCP on Mac OS X is the additional
dependency on the developer tools (XCode package). The specific need
PCP has is to have a C preprocessor available for managing the metric
namespace, which on Mac OS X is a frontend to gcc and not installed by
default.
Once this is installed, installing PCP from the DMG file is as simple
as clicking on the icon in a Finder window, and following the prompts
from the installer.
== A.3 AIX Installation
At this stage, noone is making available pre-built AIX packages.
A port to AIX has been done, and merged, however - building from
the source is currently the only option. The packaging work is also
begun on this platform (see the build/aix/ directory in the sources).
== A.4 Solaris Installation
Prebuild Solaris packages are available from PCP download section.
At this stage the package are distributed as SVR4 package datastream
and are build on Open Solaris 5.11.
You can install the package using 'pkgadd' command, e.g.:
# pkgadd -d pcp-X.Y.Z
During the installation the following three services are registered
with the Solaris' service management facility:
# svccfg list \*/pcp/\*
application/pcp/pmcd
application/pcp/pmlogger
application/pcp/pmie
application/pcp/pmproxy
On the new installation all services are disabled, during the upgrade
from the previous version of PCP the state of the services is
preserved.
Use of 'svcadm' command to enable or disable is preferred over explicit
invocation of the pmcd start script.
Use 'svcs' command to check the state of the services, e.g.:
# svcs -l application/pcp/pcp
fmri svc:/application/pcp/pcp:default
name Performance Co-Pilot Collector Daemon
enabled false
state disabled
next_state none
state_time 20 March 2012 11:33:27 AM EST
restarter svc:/system/svc/restarter:default
dependency require_all/none svc:/system/filesystem/local:default (online) svc:/milestone/network:default (online)
== A.5 Windows Installation
Download the native Windows version of PCP from the PCP download
section. There are two variants: .zip and .exe. The latter is
a self-installing executable, the former a simple compressed PCP
image. Run the executable, follow the prompts, and a Startup
Menu item with several PCP options will be available - standard
DOS shell or POSIX shell (with suitable environment setup), the
Perl CPAN configuration tool, links to documentation and online
PCP internet resources, and the pmchart utility.
== B. Building from source
1. Configure, build and install the package
The pcp package uses autoconf/configure and expects a GNU build
environment (your platform must at least have gmake).
If you just want to spin a .RPM, .DEB, .DMG, .EXE and/or tar
file, use the Makepkgs script in the top level directory.
This will configure and build the package for your platform and
leave binary and src packages in the build/<pkg-type> directory.
It will also leave binary and source tar file in the build/tar
directory.
$ ./Makepkgs --verbose
2. If you want to build the package and install it manually you will
first need to ensure the "user" pcp is created so that key parts
of the PCP installation can run as a user other than root.
For Debian this means the following (equivalent commands are
available on all distributions):
$ su root
# groupadd -r pcp
# useradd -c "Performance Co-Pilot" -g pcp -d /var/lib/pcp -M -r -s /usr/sbin/nologin pcp
Then use the following steps (use configure options to suit your
preferences, refer to the qa/admin/myconfigure script for some
guidance and see also section D below for additional details):
$ ./configure --prefix=/usr --libexecdir=/usr/lib --sysconfdir=/etc \
--localstatedir=/var --with-rcdir=/etc/init.d
$ make
$ su root
# make install
Note 0: PCP services run as non-root by default. Create unprivileged
users "pcp" with home directory /var/lib/pcp, and "pcpqa" with home
directory such as /var/lib/pcp/testsuite, or as appropriate, or
designate other userids in the pcp.conf file.
Note 1: that there are so many "install" variants out there that we
wrote our own script (see "install-sh" in the top level directory),
which works on every platform supported by PCP.
Note 2: the Windows build is particularly involved to setup, this
is primarily due to build tools not being available by default on
that platform. See the PCP Glider scripts and notes in the pcpweb
tree to configure your environment before attempting to build from
source under Win32.
== C. Post-install steps
You will need to start the PCP Collection Daemon (PMCD), as root:
Linux, AIX:
# service pmcd start (or...)
# /etc/init.d/pmcd start (or...)
# /etc/rc.d/init.d/pmcd start
Mac OS X:
# /Library/StartupItems/pcp/pmcd start
Windows:
$PCP_DIR/etc/pmcd start
Solaris:
# svcadm enable application/pcp/pmcd
Once you have started the PMCD daemon, you can list all performance
metrics using the pminfo(1) command, E.g.
# pminfo -fmdt (you don't have to be root for this, but you may need to
type rehash so your shell finds the pminfo command).
If you are writing scripts, you may find the output from pmprobe(1)
easier to parse than that for pminfo(1). There are numerous other
PCP client tools included.
PCP can be configured to automatically log certain performance metrics
for one or more hosts. The scripts to do this are documented in
pmlogger_check(1). By default this facility is not enabled. If you want
to use it, you need to
# determine which metrics to log and how often you need them
# edit $PCP_SYSCONF_DIR/pmlogger/control
# edit $PCP_SYSCONF_DIR/pmlogger/config.default
# (and any others in same dir)
# as root, "crontab -e" and add something like:
# -- typical PCP log management crontab entries
# daily processing of pmlogger archives and pmie logs
10 0 * * * $PCP_BINADM_DIR/pmlogger_daily
15 0 * * * $PCP_BINADM_DIR/pmie_daily
#
# every 30 minutes, check pmlogger and pmie instances are running
25,40 * * * * $PCP_BINADM_DIR/pmlogger_check
5,55 * * * * $PCP_BINADM_DIR/pmie_check
The pmie (Performance Metrics Inference Engine) daemon is _not_
configured to start by default. To enable it, you may want to (on
Linux platforms with chkconfig).
# su root
# chkconfig pmie on
# edit the pmie control file (usually below $PCP_SYSCONF_DIR/pmie)
# edit the config file (usually $PCP_SYSCONF_DIR/pmie/config.default)
# set up cron scripts similar to those for pmlogger (see above)
Configure some optional Performance Metrics Domain Agents (PMDAs)
The default installation gives you the metrics for cpu, per-process,
file system, swap, network, disk, memory, interrupts, nfs/rpc and
others. These metrics are handled using the platform PMDA - namely
pmda_linux.so (Linux), pmda_darwin.dylib (Mac), or pmda_windows.dll
(Windows). It also gives you the PMCD PMDA, which contains metrics
that monitor PCP itself.
There are many other optional PMDAs that you can configure, depending
on which performance metrics you need to monitor, as follows:
Note: $PCP_PMDAS_DIR is normally /var/pcp/pmdas, see pcp.conf(4).
Web Server metrics
# su root
# cd $PCP_PMDAS_DIR/apache (i.e. cd /var/pcp/pmdas/apache)
# ./Install
# Check everything is working OK
# pminfo -fmdt apache
Other PMDAs in the pcp package include:
apache - monitor apache web server stats
cisco - monitor Cisco router stats
dbping - query any database, extract response times
elasticsearch - monitor an elasticsearch cluster
kvm - monitor kernel-based virtual machine stats
mailq - monitor the mail queue
memcache - monitor memcache server stats
mmv - export memory-mapped value stats from an application
mounts - keep track of mounted file systems
mysql - monitor MySQL relational databases
postgres - monitor PostGreSQL relational databases
process - keep an eye on critical processes/daemons
roomtemp - monitor room temp (needs suitable probe)
rsyslog - monitor the reliable system log daemon
sendmail - monitor sendmail statistics
shping - ping critical system services, extract response times
trace - for instrumenting arbitrary applications, see pmtrace(1)
txmon - transaction and QOS monitoring
sample - for testing
simple - example src code if you want to write a new PMDA
trivial - even easier src code for a new PMDA.
The procedure for configuring all of these is to change to the
directory for the PMDA (usually below /var/lib/pcp/pmdas), and then
run the ./Install script found therein. None of these PMDAs are
configured by default - you choose the PMDAs you need and run the
Install script. Installation can be automated (defaults chosen) by
touching .NeedInstall in the appropriate pmdas directory and then
restarting the pmcd service via its startup script.
== D. Non-default build, install and run
To run build and run a version of PCP that is installed in a private
location (and does not require root privileges), first create the
pcp "user" as described in section B.2 above), then
$ ./configure --prefix=/some/path
This will populate /some/path with a full PCP installation. To use this
ensure the following are set in the environment:
$ export PCP_DIR=/some/path
Amend your shell's $PATH to include the PCP directories, found as
follows:
$ cd /some/path
$ xtra=`grep '^PCP_BIN' etc/pcp.conf | sed -e 's/.*=//' | paste -s -d :`
$ PATH=$xtra:$PATH
Ensure the new libraries can be found:
$ export LD_LIBRARY_PATH=`grep '^PCP_LIB' etc/pcp.conf \
| sed -e 's/.*=//' | uniq | paste -s -d :`
Tell Perl where to find loadable modules:
$ export PERL5LIB=$PCP_DIR/usr/lib/perl5:$PCP_DIR/usr/share/perl5
Allow man(1) to find the PCP manual pages:
$ export MANPATH=`manpath`:$PCP_DIR/usr/share/man
If your version is co-exiting with a running PCP in a default
install, then alternative port numbers in your environment for pmcd
($PMCD_PORT), pmlogger ($PMLOGGER_PORT) and pmproxy ($PMPROXY_PORT)
|