Merge branch 'master' of github.com:performancecopilot/pcp

build/rpm/fedora.spec

@@ -2698,8 +2698,6 @@ pmieconf -c enable dmthin
     systemctl restart pmie >/dev/null 2>&1
     systemctl enable pmcd >/dev/null 2>&1
     systemctl enable pmlogger >/dev/null 2>&1
-    systemctl enable pmlogger_daily_report >/dev/null 2>&1
-    systemctl enable pmlogger_daily_report-poll >/dev/null 2>&1
     systemctl enable pmie >/dev/null 2>&1
 %else
     /sbin/chkconfig --add pmcd >/dev/null 2>&1

build/rpm/pcp.spec.in

@@ -2780,8 +2780,6 @@ pmieconf -c enable dmthin
     systemctl restart pmie >/dev/null 2>&1
     systemctl enable pmcd >/dev/null 2>&1
     systemctl enable pmlogger >/dev/null 2>&1
-    systemctl enable pmlogger_daily_report >/dev/null 2>&1
-    systemctl enable pmlogger_daily_report-poll >/dev/null 2>&1
     systemctl enable pmie >/dev/null 2>&1
 %else
     /sbin/chkconfig --add pmcd >/dev/null 2>&1

man/man1/pmlogger_daily_report.1

@@ -30,11 +30,8 @@
 .SH DESCRIPTION
 .B pmlogger_daily_report
 and the associated
-.BR crontab (5)
-entry (depending on local platform, either 
-.I $PCP_ETC_DIR/cron.d/pcp-pmlogger-daily-report
-or
-.IR $PCP_SYSCONF_DIR/pmlogger/crontab-daily-report )
+.BR systemd (1)
+services
 write daily performance summary reports, much like those produced by 
 .BR sadc (1)
 and the
@@ -43,10 +40,11 @@ utility.
 .PP
 All of the command line arguments are optional and intended to be self
 explanatory.
-If no arguments are specified,
+The service is not enabled by default.
+If the service is enabled and no arguments are specified,
 .B pmlogger_daily_report
 will be run by
-.BR cron (8)
+.BR systemd
 at 2am each morning and write a performance summary report named
 .BI sar XX
 (where
@@ -64,11 +62,6 @@ directory by default, but this may be changed with the
 .B \-o
 option to a different
 .IR directory .
-Note the default
-.BR crontab (5)
-entry currently specifies
-.B $PCP_LOG_DIR/sa
-as the output directory; this may be changed in the future.
 .PP
 If the
 .B \-a
@@ -101,9 +94,6 @@ Note that there are suffciently flexible command line options for
 to be used to read any
 .B archivefile
 and write the report to any output directory.
-As such, this tool can be configured by editing the crontab entry,
-.B $PCP_ETC_DIR/cron.d/pcp-pmlogger-daily-report
-and adding whatever command line options are required.
 .PP
 If the
 .B \-p
@@ -157,8 +147,8 @@ capture the required archives needed by
 .BR pmlogger_daily_report .
 .PP
 In order to ensure that mail is not unintentionally sent when this
-script is run from
-.BR cron (8)
+script is run by
+.BR systemd (1),
 diagnostics are always sent to a log file.
 By default, this file is
 .B $PCP_LOG_DIR/pmlogger/pmlogger_daily_report.log
@@ -171,9 +161,7 @@ renamed with a
 suffix (overwriting any log file saved earlier) before diagnostics
 are generated to the log file.
 .PP
-The output from the
-.BR cron (8)
-execution of the script may be extended using the
+The output from the execution of the script may be extended using the
 .B \-V
 option which enables verbose tracing of activity.
 By default the script generates no log output unless some error or
@@ -198,7 +186,6 @@ as described in
 .BR pmlogger (1),
 .BR pmrep (1),
 .BR sadc (1),
-.BR crontab (5),
 .BR sa2 (8)
 and
-.BR cron (8).
+.BR systemd (1).

src/pmdas/bcc/README.md

@@ -8,59 +8,76 @@ eBPF was [described by Ingo Molnár](https://lkml.org/lkml/2015/4/14/232) as:
 
 > One of the more interesting features in this cycle is the ability to attach
 > eBPF programs (user-defined, sandboxed bytecode executed by the kernel) to
-> kprobes. This allows user-defined instrumentation on a live kernel image that
-> can never crash, hang or interfere with the kernel negatively.
+> kprobes. This allows user-defined instrumentation on a live kernel image
+> that can never crash, hang or interfere with the kernel negatively.
 
 [BCC](https://github.com/iovisor/bcc) has made creating new eBPF programs
 easier, and the BCC project offers a wide variety of such tools
-(https://github.com/iovisor/bcc/tree/master/tools). However, basically all these
-programs are individual, disjoint utilities that are mostly meant for
-interactive use. This is not a suitable approach to collect, monitor and analyze
-performance data in larger environments where there are hundreds, if not
-thousands, installations and where human intervention is unfeasible at best.
-
-This PMDA loads and acts as a bridge for any number of configured, separate BCC
-PMDA Python modules running eBPF programs. Each module consists of an eBPF
+(https://github.com/iovisor/bcc/tree/master/tools). However, basically all
+these programs are individual, disjoint utilities that are mostly meant for
+interactive use. This is not a suitable approach to collect, monitor and
+analyze performance data in larger environments where there are hundreds, if
+not thousands, installations and where human intervention is unfeasible at
+best.
+
+This PMDA loads and acts as a bridge for any number of configured, separate
+BCC PMDA Python modules running eBPF programs. Each module consists of an eBPF
 program (e.g. [sysfork.bpf](modules/sysfork.bpf)) and a corresponding Python
 file (e.g. [sysfork.python](modules/sysfork.python)).
 
 # Requirements
-* A linux kernel with eBPF support (RHEL 7.6+ or Kernel 4.7+)
+
+* A Linux kernel with eBPF support (RHEL 7.6+ or kernel 4.7+)
+* BCC (https://github.com/iovisor/bcc) and BCC Python module (0.5+)
 
 # Configuration
 
+There are several modules available to trace system and application behavior
+from different angles, including number of (failed) system calls (per pid),
+any available kernel tracepoints, USDT/dtrace/stap probes, uprobes, kprobes,
+and application internals (like new threads, objects, method calls).
+
 Modules can be enabled and configured in the [bcc.conf](bcc.conf) file. The
 `modules` setting of the `[pmda]` section lists the enabled modules:
 ```
 [pmda]
 modules = runqlat,tcplife
 ```
 
-Each module has their own section with individual module-related configuration,
-for example the `tcplife` module:
+Each module has their own section with individual module-related
+configuration, for example the `tcplife` module:
 ```
 [tcplife]
 module = tcplife
 cluster = 3
 dport = 80,443
 ```
 
-In this case the tcplife module traces only TCP sessions with a remote port
-of 80 or 443.
+In this case the tcplife module traces only TCP sessions with a remote port of
+80 or 443.
+
+For many modules process(es) to monitor can be defined as a list of names,
+PIDs, or regular expressions. PMDA-wide parameter `process_refresh` can be
+used to monitor newly created processes matching the list of processes.
 
 # Installation
+
 ```
 yum install pcp pcp-pmda-bcc
 systemctl enable --now pmcd
 cd $PCP_PMDAS_DIR/bcc
 ./Install
 ```
 
+On many systems `$PCP_PMDAS_DIR` is `/var/lib/pcp/pmdas`.
+
 # Uninstalling the PMDA
+
 ```
 cd $PCP_PMDAS_DIR/bcc
 ./Remove
 ```
+
 # Frontend
 
 ## CLI
@@ -86,23 +103,27 @@ Most of the BCC PMDA Modules have a corresponding Vector widget
 for example:
 
 ### tcplife
+
 Shows information about recently closed TCP sessions:
 
 ![vector_tcplife](https://user-images.githubusercontent.com/538011/41207752-4e216ca2-6d1b-11e8-89c6-c34a42c62351.png)
 
 ### profile
-Records stack traces at a specific interval, which will be rendered as a flamegraph in Vector:
 
-![profile](https://user-images.githubusercontent.com/538011/42831897-de7c8fca-89ef-11e8-9d35-59a89248d83c.png)
+Records stack traces at a specific interval, which will be rendered as a
+flamegraph in Vector:
 
+![profile](https://user-images.githubusercontent.com/538011/42831897-de7c8fca-89ef-11e8-9d35-59a89248d83c.png)
 
 # Troubleshooting
+
 * Check logfile for errors: `$PCP_LOG_DIR/pmcd/bcc.log`
 * Check if the bcc Python module is installed: `/usr/bin/pcp python -c 'import bcc'`
 * Check if BCC PMDA metrics are registered: `pminfo | grep bcc`
 * Check if BCC PMDA metrics contain data: `pminfo -f bcc.proc.sysfork` (if the `sysfork` module is enabled)
-* Check if BCC is working on this system: `yum install bcc-tools && /usr/share/bcc/tools/runqlat`
+* Check if BCC is working on the system: `yum install bcc-tools && /usr/share/bcc/tools/runqlat`
 
 # License
+
 * PMDA: [GPLv2](https://github.com/performancecopilot/pcp/blob/master/COPYING)
 * eBPF/BCC programs: [Apache License 2.0](https://github.com/iovisor/bcc/blob/master/LICENSE.txt)