|
HTML rendering created 2025-02-02 by Michael Kerrisk, author of The Linux Programming Interface. For details of in-depth Linux/UNIX system programming training courses that I teach, look here. Hosting by jambit GmbH. |
|
|
NAME | SYNOPSIS | DESCRIPTION | FILES | PCP ENVIRONMENT | SEE ALSO | COLOPHON |
|
|
|
WEBVIS(1) General Commands Manual WEBVIS(1)
webvis - visualize system-level Web server activity
webvis [-CVz] [-A align] [-a archive] [-b maxbusy] [-h host] [-i
maxio] [-m max] [-n pmnsfile] [-O time] [-p port] [-r maxreq] [-S
time] [-T time] [-t interval] [-x version] [-Z timezone]
[interface ...]
webvis displays an overview of system level Web server performance
statistics collected from the Performance Co-Pilot (PCP)
infrastructure. The display is modulated by the values of the
performance metrics retrieved from the target host (which is
running pmcd(1) and the pmdaweblog(1) Performance Metrics Domain
Agent) or from the PCP archive log identified by archive. The
display is updated every interval seconds (default 2 seconds).
As in all pmview(1) scenes, when the mouse is moved over one of
the bars, the current value and metric information for that bar
will be shown in the text box near the top of the display.
The height of the web request and network activity bars is
proportional to the performance metric values relative to the
maximum expected activity, as controlled by the -m and -r options
(see below). Similarly the -b and -i options control the scaling
for disk activity bars.
The bars in the webvis scene represent the following information;
Requests by Size
At the front of the scene, the "Requests by Size" row of bars
shows the rate of requests for different size requests (the
histograms are defined by the following byte counts: 0, 3
Kbytes, 10 Kbytes, 30 Kbytes, 100 Kbytes, 300 Kbytes, 1 Mbyte,
3 Mbytes and larger than 3 Mbytes). Notice that the size
divisions are not evenly distributed. The "size" is the data
portion of the response to each Web server request. These
rates are aggregated across all monitored Web servers.
Requests by Type
This row of bars shows the request rate for each type of HTTP
request (get, post, head and other), aggregated across all
monitored Web servers. For a detailed display showing the
break down of requests per Web server, see weblogvis(1).
Network
For every network interface there are two stacked bars. One
of the bars shows the input traffic while the other bar shows
the output traffic. The stacks are composed of the number of
errors (red), the number of drops (orange) and the number of
packets (green). In general, if there are any "dropped input
packets" then the corresponding network interface is
saturated, or there are insufficient network resources
available in the kernel to adequately service the input
request load. If this is the case then the Alarm Conditions
rows (see below) may provide more detail into the source of
the problem.
Alarm Conditions
The red row of bars shows an assortment of TCP error
conditions (aggregated for all network interfaces), the orange
bars show critical kernel buffer allocation problems, and the
yellow bar shows severe paging conditions. If any of these
bars have a non-zero height then the system being monitored
may require kernel parameter tuning, software reconfiguration
or more hardware resources. The performance metrics behind
the bars are:
network.tcp.drops
- rate of dropped connections
network.tcp.conndrops
- rate of embryonic connections dropped
network.tcp.timeoutdrop
- rate of connections dropped by rexmit timeout
network.tcp.rcvbadsum
- rate of packets discarded for bad checksums
network.tcp.rexmttimeo
- rate of retransmit timeouts
network.tcp.sndrexmitpack
- rate of data packets retransmitted
swap.pagesout
- page swap out rate (indicating insufficient memory)
network.mbuf.failed
- rate of incidents where the kernel failed to find
mbuf space
network.mbuf.waited
- rate of incidents where the kernel waited to find
mbuf space
CPU This column shows CPU utilization, aggregated over all CPUs.
(CPU idle time is not included in the column).
Disk
There are two cylinders showing disk metrics. The first
cylinder shows the rate of read (yellow) and write (violet)
operations, aggregated over all disk spindles. The second
cylinder shows the average (over all disks) percentage of time
for which a disk is busy or active. This metric is not
available in PCP1.x versions, therefore if webvis is being
used to monitor a host running PCP1.x this cylinder will not
be displayed.
To adjust the scaling of these objects, refer to the -b and -i
options described below.
Mem There are two bars showing memory metrics. The first bar
shows utilized memory, with different colors representing
different types of utilization (kernel, user, etc), while the
second bar shows the amount of free memory. If webvis is
being used to monitor a host running PCP1.x then only the bar
showing free memory will be displayed.
If any optional interface arguments are specified in the command
line, then just the network interfaces matching the interface
arguments will appear in the Network section. By default, all
interfaces will be used. The interface arguments are used as
patterns for egrep(1) matching against the interface names, so ec
would select all external Ethernet interfaces for a Challenge S.
webvis uses pmview(1), and so the user interface follows that
described for pmview(1), which in turn displays the scene within
an Inventor examiner viewer.
webvis passes most command line options to pmview(1). Therefore,
the command line options -A, -a, -C, -h, -n, -O, -p, -S, -t, -T,
-x, -Z and -z, and the user interface are described in the
pmview(1) man page.
Options specific to webvis are:
-b maxbusy
Controls the maximum (normalization) value for the average
percentage of the time active over all disks. The default
value is 30% active.
-i maxio
Controls the maximum (normalization) value for the sume of
the aggregate disk read and disk write rates. The default
value is 100 I/Os per second.
-m max Controls the maximum (normalization) value for the packet
input and packet output rates. The default value is 750
packets/second.
-r maxreq
Controls the maximum Web request rate. The default is 5%
of the maximum packet rate (i.e. 38 requests/second by
default). The maximum Web error rate is fixed at 20% of
the maximum Web request rate (i.e. 7 errors/second by
default).
-V The derived configuration file for pmview(1) is written on
standard output. This may be saved and used directly with
pmview if the user wishes to customize the display, or
modify some of the normalization parameters.
$PCP_VAR_DIR/pmns/*
default PMNS specification files
$PCP_VAR_DIR/config/pmlogger/config.web
pmlogger(1) configuration file that can be used to create a
PCP archive suitable for display with webvis
Environment variables with the prefix PCP_ are used to
parameterize the file and directory names used by PCP. On each
installation, the file /etc/pcp.conf contains the local values for
these variables. The $PCP_CONF variable may be used to specify an
alternative configuration file, as described in pcp.conf(4).
pmcd(1), pmchart(1), pmdaweblog(1), pmdawebping(1), pmdumplog(1),
pminfo(1), pmlogger(1), pmval(1), pmview(1), weblogvis(1),
webpingvis(1), pcp.conf(4) and pcp.env(4).
This page is part of the PCP (Performance Co-Pilot) project.
Information about the project can be found at
⟨http://www.pcp.io/⟩. If you have a bug report for this manual
page, send it to pcp@groups.io. This page was obtained from the
project's upstream Git repository
⟨https://github.com/performancecopilot/pcp.git⟩ on 2025-02-02.
(At that time, the date of the most recent commit that was found
in the repository was 2025-01-30.) If you discover any rendering
problems in this HTML version of the page, or you believe there is
a better or more up-to-date source for the page, or you have
corrections or improvements to the information in this COLOPHON
(which is not part of the original manual page), send a mail to
man-pages@man7.org
Performance Co-Pilot WEBVIS(1)
Pages that refer to this page: pmdaweblog(1), weblogvis(1), webpingvis(1)
|
HTML rendering created 2025-02-02 by Michael Kerrisk, author of The Linux Programming Interface. For details of in-depth Linux/UNIX system programming training courses that I teach, look here. Hosting by jambit GmbH. |
|