1 Notes on using the PCP QA Suite
2 ===============================
7 The PCP QA Suite is designed with a philosophy that it is trying to
8 exercise the code in a context that is as close as possible to that
9 which an end-user would experience. For this reason, the PCP software to
10 be tested should be installed in the "usual" places, with the "usual"
11 permissions and operate on the "usual" ports.
13 In particular the QA Suite does not execute PCP applications like pmcd,
14 pmlogger, pminfo, pmie, pmval, etc from the source tree. Rather they
15 need to have been built and packaged and installed on the local system
16 prior to starting any QA. Refer to the ../Makepkgs script for a recipe
17 that may be used to build packages for a variety of platforms.
19 Further the PCP QA Suite exercises and tests aspects of the PCP
20 packaging, use of certain local accounts, interaction with system
21 daemons, init systems, a number of PCP-related system administrative
22 functions, e.g. to stop and start PCP services. Refer to the notes
25 But this also means the QA Suite may alter existing system configuration
26 files, and this introduces some risk, so PCP QA should not be run
27 on production systems. Historically we have used developer systems
28 and dedicated QA systems for running the full QA Suite - VMs are
29 particularly well-suited to this task.
31 In addition to the base PCP package installation, the sample and simple
32 PMDAs need to be installed (however the QA infrastructure will take
33 care of this, e.g. by running ./check 0).
38 There is some local configuration needed ... check the file
39 "common.config" ... this script uses heuristics to set a number of
40 interesting variables, specifically:
43 The $DISPLAY setting for an X server that is willing to accept
44 connections from X clients running on the local machine. This is
45 optional, and if not set any QA tests dependent on this will
49 The hostname for a host running pmcd, but the host is preferably
50 a long way away (over a WAN) for timing test. This is optional,
51 and if not set any QA tests dependent on this will be skipped.
54 The hostname for a host running pmcd, with a hyphen (-) in the
55 hostname. This is optional, and if not set any QA tests dependent
56 on this will be skipped.
58 Next, mk.qa_hosts is a script that includes heuristics for selecting
59 and sorting the list of potential remote PCP QA hosts (qa_hosts.master).
60 Refer to the comments in qa_hosts.master, and make appropriate changes.
62 For each of the potential remote PCP QA hosts, the following must be
65 (a) PCP installed from packages,
67 (c) a login for the user "pcpqa" needs to be created, and then set
68 up in such a way that ssh/scp will work without the need for any
69 password, i.e. these sorts of commands
70 $ ssh pcpqa@pcp-qa-host some-command
71 $ scp some-file pcpqa@pcp-qa-host:some-dir
72 must work correctly when run from the local host. The "pcpqa"
73 user's environment must also be initialized so that their shell's
74 path includes all of the PCP binary directories (identify these
75 with $ grep BIN /etc/pcp.conf), so that all PCP commands are
76 executable without full pathnames. Of most concern would be
77 auxilliary directory (usually /usr/pcp/bin, /usr/share/pcp/bin or
78 /usr/libexec/pcp/bin) where commands like pmlogger(1), pmhostname(1),
79 mkaf(1), etc.) are installed. And finally, the "pcpqa" user needs
80 to be included in the group "pcp".
82 Once you've modified common.config and qa_hosts.master, then run
83 "chk.setup" to validate the settings.
85 For test 051 we need five local hostnames that are valid, although PCP
86 does not need to be installed there, nor pmcd(1) running. The five
87 hosts listed in 051.hosts (the comments at the start of this file
88 explain what is required) should suffice for most installations.
90 The PCP QA tests are designed to be run by a non-root user. Where root
91 privileges are needed, e.g. to stop or start pmcd, install/remove
92 PMDAs, etc. the "sudo" application is used. When using sudo for QA,
93 your current or pcpqa user needs to be able to execute commands as
94 root without being prompted for a password. This can be achieved by
95 adding the following line to the /etc/sudoers file (or in more recent
96 versions of sudo, a /etc/sudoers.d/pcpqa file):
98 pcpqa ALL=(ALL) NOPASSWD: ALL
100 Some tests are graphical, and wish to make use of your display.
101 For authentication to success, you may find you need to perform some
102 access list updates, e.g. "xhost +local:" for such tests to pass
105 You can now verify your QA setup, by running:
109 The first time you run "check" (see below) it will descend into the
110 src directory (see below) and make all of the QA test programs and
111 dynamic PCP archives, so some patience may be required.
113 If test 000 fails, it may be that you have locally developed PMDAs
114 or optional PMDAs installed. Edit common.filter, and modify the
115 _filter_top_pmns() procedure to strip the top-level name components
116 for any new metric names (there are lots of examples already there)
117 ... if these are distributed (shipped) PMDAs, please update the list.
119 Firewalls can get in the way. In addition to the standard pmcd port(s)
120 (TCP ports 44321, 44322 and 44323) one needs to open ports to allow
121 incoming connections and outgoing connections on a range of ports
122 for pmdatrace, pmlogger connections via pmlc, and some QA tests.
123 Opening the TCP range 4320 to 4350 (inclusive) should suffice.
125 If the avahi services are to be tested, then the firewall also needs
126 to allow mDNS traffic (UDP, port 5353), for both external and internal
134 This script runs tests and verifies the output. In general, test NNN
135 is expected to terminate with an exit status of 0, no core file and
136 produce output that matches that in the file NNN.out ... failures
137 leave the current output in NNN.out.bad, and may leave a more
138 verbose trace that is useful for diagnosing failures in NNN.full.
140 The command line options to check are:
142 NNN run test NNN (leading zeros will be added as necessary to
143 the test sequence number, so 00N and N are equivalent)
145 NNN- all tests >= NNN
147 NNN-MMM all tests in the range NNN ... MMM
149 -l diffs in line mode (the default is to use xdiff or similar)
151 -n show me, do not run any tests
153 -q quick mode, by-pass the initial setup integrity checks
154 (recommended that you do not use this the first time, nor
155 if the last run test failed)
157 -g xxx include tests from a named group (xxx) ... refer to the
160 -x xxx exclude tests from a named group (xxx) ... refer to the
163 If none of the NNN variants or -g is specified, then the default
166 Each of the NNN scripts that may be run by check follows the same
169 - include some optional shell procedures and set variables to
170 define the local configuration options
171 - optionally, check the run-time environment to see if it makes
172 sense to run the test at all, and if not echo the reason to the
173 file NNN.notrun and exit ... check will notice the NNN.notrun
174 file and skip any testing of the exit status or comparison
176 - define $tmp as a prefix to be used for all temporary files, and
177 install a trap handler to remove temporary files when the scipt
179 - optionally, check the run-time environment to choose one of
180 a number of expected output formats, and link the selected
181 file to NNN.out ... if the same output is expected in all
182 environments, the NNN.out file will already exist as part of
183 the PCP QA distribution
185 - optionally save all the output in the file NNN.full ... this
186 is only useful for debugging test failures
187 - filter the output to produce deterministic output that will
188 match NNN.out if the test has been successful
191 This script creates a new NNN.out file. Since the NNN.out files
192 are precious, and reflect the state of the qualified and expected
193 output, they should typically not be changed unless some change
194 has been made to the NNN script or the filters it uses.
197 Make sure "group" is writeable, then run "new" to create the
198 skeletal framework of a new test.
200 It is strongly suggested that you base your test on an existing test
201 ... pay particular attention to making the output deterministic
202 so the test uses the "not run" protocols (see 009 and check for
203 examples) to avoid running the test (and hence failing) if an
204 optional application, feature or platform is not available, and
205 uses appropriate filters (see common.filter for lots of useful
206 filters already packaged as shell procedures).
209 Report differences between the NNN.out and NNN.out.bad files.
210 By default, uses all of the NNN.out.bad files in the current
211 directory, but can also specify test numbers or ranges of test
212 numbers on the command line.
214 Other options may be used to fetch good and bad output files from
215 various exotic remote locations (refer to the script).
218 Make in the src Directory
219 -----------------------------
221 The src directory contains a number of test applications that are
222 designed to exercise some of the more exotic corners of the PCP
225 In making these applications, you may see this ...
227 Error: trace_dev.h and ../../src/include/trace_dev.h are different!
228 make: [trace_dev.h] Error 1 (ignored)
230 this is caused by the source for the pcp_trace library being out of sync
231 with the src applications. If this happens, please ...
234 2. diff -u trace_dev.h ../../src/include/trace_dev.h
235 and mail the differences to pcp@oss.sgi.com so we can refine the
236 Makefiles to avoid cosmetic differences
237 3. mv trace_dev.h trace_dev.h.orig
238 cp ../../src/include/trace_dev.h trace_dev.h
245 Test 008 depends on the local disk configuration, so you need to
246 make your own 008.out file (or rather a variant that 008 will link to
247 008.out when the test is run). Refer to the 008 script, but here is
250 $ touch touch 008.out.`hostname`
252 $ mv 008.out 008.out.`hostname`
254 Be aware that it can be adversely influenced by temporary disks like
255 USB sticks, mobile phones, or other transient storage that may come
256 and go in your test systems.
262 If you find something that does not work, and fix it, or create
263 additional QA tests, please send the details to pcp@oss.sgi.com.