Ubuntu Manpages

pmproxy

proxy for performance metrics collector and querying

pmproxy [-AdfFt?] [-c conffile] [-D debug] [-h host[,host ...] [-i ipaddress] [-l logfile] [-L bytes] [-p port[,port ...] [-r port[,port ...] [-s sockname] [-U username] [-x outfile]

pmproxy acts as a protocol proxy, allowing Performance Co-Pilot (PCP) monitoring clients to connect to one or more pmcd(1) and/or key-value servers (such as https://valkey.io/) indirectly.

In its default mode of operation pmproxy provides the REST API for PCP services (see PMWEBAPI(3) for details). This includes provision of an Open Metrics - https://openmetrics.io - text interface for PCP metrics at /metrics, real-time access to PCP metrics through the /pmapi interfaces, and access to the fast, scalable PCP time series query capabilities offered in conjunction with a key-value server (see pmseries(1) for details) via the /query REST interfaces.

pmproxy can be deployed in a firewall domain, or on a cluster ``head'' node where the IP (Internet Protocol) address of the hosts where pmcd and/or a key-value server (such as https://valkey.io/) are running may be unknown to the PCP monitoring clients, but where the IP address of the host running pmproxy is known to these clients. Similarly, the clients may have network connectivity only to the host where pmproxy is running, while there is network connectivity from that host to the hosts of interest where pmcd and/or a key-value server are running.

The behaviour of the PCP monitoring clients is controlled by either the PMPROXY_HOST environment variable or through the extended hostname specification (see PCPIntro(1) for details). If neither of these mechanisms is used, clients will make their PMAPI(3) connections directly to pmcd. If the proxy hostname syntax is used or PMPROXY_HOST is set, then this should be the hostname or IP address of the system where pmproxy is running, and the clients will connect to pmcd or a key-value server indirectly through the protocol proxy services of pmproxy.

The available command line options are:

Disable service advertisement. By default, pmproxy will advertise its presence on the network using any available mechanisms, assisting remote monitoring tools with finding it. These mechanisms are disabled with this option.
Specify the path to an optional configuration conffile, with format as described in the ``CONFIGURATION'' section. This option implies pmproxy is running in timeseries mode.
By default pmproxy prefers to run in the new timeseries mode, providing REST APIs, asynchronous network I/O, scalable time series, and secure connections using OpenSSL. However, legacy deployments may wish to use the original synchronous pmproxy implementation using libpcp networking; this can be achieved using this option. Note that the -d and -t options are mutually exclusive.
By default pmproxy is started as a daemon. The -f option indicates that it should run in the foreground. This is most useful when trying to diagnose problems with establishing connections.
Like -f, the -F option runs pmproxy in the foreground, but also does some housekeeping (like create a ``pid'' file and change user id). This is intended for use when pmproxy is launched from systemd(1) or launchd(1) and the daemonising has already been done by systemd(1) or launchd(1) and does not need to be done again by pmproxy, which is the case when neither -f nor -F is specified.

At most one of -f and -F may be specified.

Specify an alternate key-value server host to connect to for time series querying, overriding any configuration file settings. This option implies pmproxy is running in timeseries mode.
This option is usually only used on hosts with more than one network interface (very common for firewall and ``head'' node hosts where pmproxy is likely to be deployed to arbitrate access to an internal network). If no -i options are specified pmproxy accepts PCP client connections on any of its host's IP addresses. The -i option is used to specify explicitly an IP address that PCP client connections should be accepted on. ipaddress should be in the standard dotted form (e.g. 100.23.45.6). The -i option may be used multiple times to define a list of IP addresses. When one or more -i options is specified, attempted connections made on any other IP addresses will be refused.
By default a log file named pmproxy.log is written in the current directory. The -l option causes the log file to be written to a given logfile instead of the default. If this logfile cannot be created or is not writable, output is written to the standard error instead.
PDUs received by pmproxy from PCP monitoring clients are restricted to a maximum size of 65536 bytes by default to defend against Denial of Service attacks. The -L option may be used to change the maximum incoming PDU size.
Specify an alternate port number to listen on for client connections. The default value is 44322.
Require that all client connections provide a trusted client certificate.
Specify an alternate key-value server port number to connect to for time series querying, overriding any configuration file settings. This option implies pmproxy is running in timeseries mode.
Specify the path to a local unix domain socket (for platforms supporting this socket family only). The default value is $PCP_RUN_DIR/pmproxy.socket. This option implies pmproxy is running in timeseries mode.
Require that all client connections be authenticated.
Operate in automatic archive timeseries discovery mode. This mode of operation will enable the PMWEBAPI(3) REST APIs, dynamically and automatically detect active system archives being written by pmlogger(1) and import them into a key-value server (such as https://valkey.io/), for fast, scalable time series querying described in pmseries(1). Note that in this mode of operation, pmproxy only "log-tails" and ingests actively growing archives, e.g. as written by one or more pmlogger(1) instances. When an archive is first discovered (usually but not limited to pmproxy startup), all metadata is loaded and sent to the configured key-value server however note that only new archive metric value data from the tail end of each archive is ingested. Compressed archives never grow and so are ignored. See the --load option to pmseries(1) for a supported mechanism for manually loading all of the metric value data from previously collected (inactive) archives, whether compressed or not. It would be normal, though not mandated, for a set of archives being manually loaded to cover the same time period, e.g. archive data for a particular week for one or more hosts in the same data-centre.
Assume the identity of the given username before starting to accept incoming packets from PCP monitoring clients.
Before the pmproxy logfile can be opened, pmproxy may encounter a fatal error which prevents it from starting. By default the output describing this error is sent to /dev/tty but it may redirected to outfile.
-?, --help
Display usage message and exit.

When running in the timeseries mode of operation, runtime configuration is relatively complex and typically handled via the $PCP_SYSCONF_DIR/pmproxy/pmproxy.conf file. This file is in the common ``ini'' format, with section headers and individual variables and values with each section. The configuration file installed as part of PCP documents every available section and option.

At a high level, the [pmproxy] section can be used to explicitly enable or disable each of the different protocols.

The [http] section provides fine-tuning over HTTP server settings used by pmproxy. chunksize sets the chunked transfer encoding buffer size, and defaults to the system pagesize. Access control HTTP protocol settings can be adjusted using the Access-Control-Allow-Headers and Access-Control-Max-Age options. Discussion of these HTTP protocol headers is beyond the scope of this document, but suitable default values are described within the pmproxy.conf configuration file.

The [keys] section allows connection information for one or more backing key-value server processes to be configured (hostnames and ports). Note to access multiple (scalable) key-value servers, the servers variable in this section can be a comma-separated list of hostname:port pairs. Alternatively, it can be a single key-value server host that will be queried using the "CLUSTER INFO" command to automatically configure multiple backing hosts.

In earlier versions of PCP (before 6) an alternative configuration setting section was used for this purpose - key-value servers were specified in the [pmseries] section and this is still accepted as a fallback for backwards compatibility.

Normally, pmproxy is started automatically at boot time and stopped when the system is being brought down. Under certain circumstances it is necessary to start or stop pmproxy manually. To do this one must become superuser and type

# $PCP_RC_DIR/pmproxy start

to start pmproxy, or

# $PCP_RC_DIR/pmproxy stop

to stop pmproxy. Starting pmproxy when it is already running is the same as stopping it and then starting it again.

Normally pmproxy listens for PCP client connections on TCP/IP port number 44322 (as well as 44323 with timeseries enabled) registered at https://www.iana.org/. Either the environment variable PMPROXY_PORT or the -p command line option may be used to specify alternative port number(s) when pmproxy is started; in each case, the specification is a comma-separated list of one or more numerical port numbers. Should both methods be used or multiple -p options appear on the command line, pmproxy will listen on the union of the set of ports specified via all -p options and the PMPROXY_PORT environment variable. If non-default ports are used with pmproxy care should be taken to ensure that PMPROXY_PORT is also set in the environment of any client application that will connect to pmproxy, or that the extended host specification syntax is used (see PCPIntro(1) for details).

If pmproxy is already running the message "Error: OpenRequestSocket bind: Address already in use" will appear. This may also appear if pmproxy was shutdown with an outstanding request from a client. In this case, a request socket has been left in the TIME_WAIT state and until the system closes it down (after some timeout period) it will not be possible to run pmproxy.

In addition to the standard PCP debugging options, see pmdbg(1), pmproxy currently supports the debugging option context for tracing client connections and disconnections.

$PCP_PMPROXYOPTIONS_PATH
command line options for pmproxy when launched from $PCP_RC_DIR/pmproxy All the command line option lines should start with a hyphen as the first character.
$PCP_SYSCONFIG_DIR/pmproxy
Environment variables that will be set when pmproxy executes. Only settings of the form "PMPROXY_VARIABLE=value" will be honoured.
./pmproxy.log
(or $PCP_LOG_DIR/pmproxy/pmproxy.log when started automatically)
All messages and diagnostics are directed here
/etc/pki/tls
default OpenSSL certificate database directory, optionally used for Secure Socket Layer connection in timeseries mode of operation. These certificates can be created and queried using the openssl tool, amongst others.

In addition to the PCP environment variables described in the PCP ENVIRONMENT section below, there are several environment variables that influence the interactions between a PCP monitoring client, pmproxy and pmcd.

For the PCP monitoring client this (or the default port number) is passed to pmproxy and used to connect to pmcd. In the environment of pmproxy PMCD_PORT is not used.
For the PCP monitoring client this is the hostname or IP address of the host where pmproxy is running. In recent versions of PCP (since version 3) this has been superseded by the extended hostname syntax (see PCPIntro(1) for details).
For the PCP monitoring client this is the port on which pmproxy will accept connections. The default is 44322, as well as 44323 with timeseries enabled.
(see PCPIntro(1)) For the PCP monitoring client, setting these environment variables will modify the timeouts used for interactions between the client and pmproxy (independent of which pmcd is being used). For pmproxy these same environment variables control the timeouts between pmproxy and all pmcd(1) instances (independent of which monitoring client is involved).

If set to the value 1, the PMPROXY_LOCAL environment variable will cause pmproxy to run in a localhost-only mode of operation, where it binds only to the loopback interface.

The PMPROXY_MAXPENDING variable can be set to indicate the maximum length to which the queue of pending client connections may grow.

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(5).

For environment variables affecting PCP tools, see pmGetOptions(3).

The -D or --debug option enables the output of additional diagnostics on stderr to help triage problems, although the information is sometimes cryptic and primarily intended to provide guidance for developers rather than for end-users. debug is a comma separated list of debugging options; use pmdbg(1) with the -l option to obtain a list of the available debugging options and their meaning.

Debugging options specific to pmproxy are as follows:

Option Description
appl0 client connections and disconnections
appl1 desperate logging mode, where a period followed by the PID of pmproxy is inserted in the name of Rlogfile before the last period, so for example pmproxy.log becomes pmproxy.<pid>.log
appl2 log incoming HTTP URLs (this is also enabled by the http debugging option, but the latter has broader scope because it turns on debugging in the libraries that pmproxy uses)

PCPIntro(1), pmcd(1), pmdbg(1), pmlogger(1), pmseries(1), PMAPI(3), PMWEBAPI(3), pmGetOptions(3), pcp.conf(5) and pcp.env(5).