Merge pull request #1942 from natoscott/postgres-pmcheck

pmcheck: improve redis, add postgresql and uwsgi checks

qa/1491

@@ -65,11 +65,21 @@ s/could be activated/OK/
 s/cannot be activated/OK/
 s/measure overhead of PCP components/.../
 s/not currently available for .*/.../
+}' \
+	-e '/^pmda-postgresql /{
+s/inactive/OK/
+s/could be activated/OK/
+s/cannot be activated/OK/
 }' \
 	-e '/^pmda-redis /{
 s/inactive/OK/
 s/could be activated/OK/
 s/cannot be activated/OK/
+}' \
+	-e '/^pmda-uwsgi /{
+s/inactive/OK/
+s/could be activated/OK/
+s/cannot be activated/OK/
 }' \
 	-e '/^zeroconf /{
 s/inactive/OK/

qa/1491.out

@@ -3,35 +3,43 @@ basic stuff ...
 -- no args --
 pmcd            OK
 pmda-overhead   OK
+pmda-postgresql OK
 pmda-redis      OK
 pmda-sample     OK
+pmda-uwsgi      OK
 pmie            OK
 pmlogger        OK
 pmproxy         OK
 zeroconf        OK
 -- -l --
 pmcd           
 pmda-overhead  
+pmda-postgresql
 pmda-redis     
 pmda-sample    
+pmda-uwsgi     
 pmie           
 pmlogger       
 pmproxy        
 zeroconf       
 -- -lv --
 pmcd            Performance Metrics Collection Daemon - local source of performance data
 pmda-overhead   overhead PMDA - ...
+pmda-postgresql PostgreSQL PMDA - metrics from PostgreSQL
 pmda-redis      Redis PMDA - metrics from redis-server(1)
 pmda-sample     sample PMDA
+pmda-uwsgi      uWSGI PMDA - metrics from uWSGI servers
 pmie            Inference Engine - rule-based monitoring
 pmlogger        Archive logger - record performance data for subsequent replay
 pmproxy         Proxy Daemon - proxy, web and redis integration
 zeroconf        Zeroconf Package ...
 -- -s --
 pmcd            OK
 pmda-overhead   OK
+pmda-postgresql OK
 pmda-redis      OK
 pmda-sample     OK
+pmda-uwsgi      OK
 pmie            OK
 pmlogger        OK
 pmproxy         OK

qa/1992

@@ -15,26 +15,39 @@ echo "QA output created by $seq"
 
 [ -d $PCP_PMDAS_DIR/uwsgi ] || _notrun "uwsgi PMDA directory is not installed"
 
+_cleanup()
+{
+    _cleanup_pmda $iam
+    $sudo rm -f $tmp.*
+    exit $status
+}
+
+iam=uwsgi
 status=1        # failure is the default!
 $sudo rm -rf $tmp $tmp.* $seq.full
+trap "_cleanup" 0 1 2 3 15
 
 pmdauwsgi_remove()
 {
     echo
-    echo "=== remove uwsgi agent ==="
+    echo "=== remove $iam agent ==="
     $sudo ./Remove >$tmp.out 2>&1
     _filter_pmda_remove <$tmp.out
 }
 
 pmdauwsgi_install()
 {
     # start from known starting points
-    cd $PCP_PMDAS_DIR/uwsgi
+    cd $PCP_PMDAS_DIR/$iam
     $sudo ./Remove >/dev/null 2>&1
     _service pmcd stop | _filter_pcp_stop
 
     echo
-    echo "=== uwsgi agent installation ==="
+    echo "=== $iam check post-install ==="
+    pmcheck pmda-$iam
+
+    echo
+    echo "=== $iam agent installation ==="
     $sudo ./Install </dev/null >$tmp.out 2>&1
     # Check metrics have appeared ... X metrics and Y values (or)
     # Check metrics have appeared ... N warnings, X metrics and 0 values
@@ -50,14 +63,15 @@ pmdauwsgi_install()
                                            }
                                          }
                                          { print }' | tr -s ' '
+    echo
+    echo "=== $iam check post-install ==="
+    pmcheck pmda-$iam
 }
 
-_prepare_pmda uwsgi
-trap "_cleanup_pmda uwsgi; exit \$status" 0 1 2 3 15
-
+# real QA test starts here
+_prepare_pmda $iam
 _stop_auto_restart pmcd
 
-# real QA test starts here
 pmdauwsgi_install
 pmdauwsgi_remove
 

src/pmdas/postgresql/.gitignore

@@ -0,0 +1,2 @@
+domain.h
+pmns

src/pmdas/postgresql/GNUmakefile

@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2018-2020 Red Hat.
+# Copyright (c) 2018-2020,2024 Red Hat.
 #
 # This program is free software; you can redistribute it and/or modify it
 # under the terms of the GNU General Public License as published by the
@@ -47,9 +47,11 @@ install_pcp install: default
 	$(INSTALL) -m 755 -d $(LOGCONFDIR)
 	$(INSTALL) -m 755 -d $(LOGCONFVARDIR)
 	$(INSTALL) -m 644 -t $(LOGCONFVARDIR)/summary pmlogconf.summary $(LOGCONFDIR)/summary
+	$(INSTALL) -m 755 $(IAM).pmcheck $(PCP_SHARE_DIR)/lib/pmcheck/pmda-$(IAM)
 else
 build-me:
-install_pcp install:
+install_pcp install: $(TOPDIR)/src/pmcheck/pmda.na.template
+	$(INSTALL) -m 755 $< $(PCP_SHARE_DIR)/lib/pmcheck/pmda-$(IAM)
 	@$(INSTALL_MAN)
 endif
 

src/pmdas/postgresql/pmdapostgresql.python

@@ -1087,6 +1087,10 @@ class POSTGRESQLPMDA(PMDA):
         # success
         return self.cur
 
+    def domain_probe(self):
+        """ return True if postgresql is available and monitorable """
+        return self.conn is not None
+
     def debug(self, msg):
         """ print diagnostic message if verbose logging is enabled """
         if self.verbose:

src/pmdas/postgresql/postgresql.pmcheck

@@ -0,0 +1,49 @@
+#!/bin/sh
+#
+# Postgresql PMDA "plugin" for pmcheck
+#
+
+. $PCP_DIR/etc/pcp.env || exit 1
+. $PCP_SHARE_DIR/lib/checkproc.sh
+
+_do_args "$@"
+
+_check()
+{
+    test -n "$@" && echo "$@" >> $tmp/out
+    [ "$verbose" -gt 0 -a -s $tmp/out ] && cat $tmp/out
+    [ $status -eq 0 ] || exit
+}
+
+iam=postgresql
+iname=postgres
+
+if $lflag
+then
+    [ "$verbose" -gt 0 ] && echo "PostgreSQL PMDA - metrics from PostgreSQL"
+elif $sflag
+then
+    status=0  # assume active until proven not to be
+    _ctl_svc state $iam || status=$?
+    _check "$iam service status: $status"
+    _ctl_pmda state $iam || status=$?
+    _check "$iam PMDA status: $status"
+    # No longer use probe due to difficulties with the PMDA implementation;
+    # in particular it switches user to postgres which we cannot easily do.
+    # Heuristic instead is simply to test to see if a postgres user exists.
+    #PROBE="$PCP_PYTHON_PROG $PCP_PMDAS_DIR/$iam/pmda$iam.python"
+    #PCP_PYTHON_PROBE=1 $PROBE || status=2
+    id $iname >/dev/null 2>&1 || status=2
+    _check "$iname user status: $status"
+elif $aflag
+then
+    _ctl_pmda activate $iam pmda$iam.python || status=1
+elif $dflag
+then
+    _ctl_pmda deactivate $iam || status=1
+else
+    [ $verbose -gt 0 ] && echo "botch sflag=$sflag aflag=$aflag dflag=$dflag show_me=$show_me verbose=$verbose"
+    status=99
+fi
+
+exit

src/pmdas/redis/redis.pmcheck

@@ -24,7 +24,9 @@ then
     which redis-server >/dev/null 2>&1 || status=2
     which redis-cli >/dev/null 2>&1 || status=2
     _check "redis install status: $status"
-    _ctl_svc state redis || status=$?
+    svc=redis
+    test -f "$PCP_SYSTEMDUNIT_DIR/redis-server.service" && svc=redis-server
+    _ctl_svc state $svc || status=$?
     _check "redis service status: $status"
     pong=`redis-cli PING`
     test "$pong" = "PONG" || status=2

src/pmdas/uwsgi/GNUmakefile

@@ -48,10 +48,12 @@ install_pcp install: default
 	$(INSTALL) -m 755 -d $(PMLOGCONFVARDIR)
 	$(INSTALL) -m 644 -t $(PMLOGCONFVARDIR)/summary pmlogconf.summary $(PMLOGCONFDIR)/summary
 	$(INSTALL) -m 644 -t $(PMDATMPDIR)/$(IAM).conf $(IAM).conf $(PMDACONFIG)/$(IAM).conf
+	$(INSTALL) -m 755 $(IAM).pmcheck $(PCP_SHARE_DIR)/lib/pmcheck/pmda-$(IAM)
 	@$(INSTALL_MAN)
 else
 build-me:
-install_pcp install:
+install_pcp install: $(TOPDIR)/src/pmcheck/pmda.na.template
+	$(INSTALL) -m 755 $< $(PCP_SHARE_DIR)/lib/pmcheck/pmda-$(IAM)
 	@$(INSTALL_MAN)
 endif
 

src/pmdas/uwsgi/pmdauwsgi.python

@@ -121,6 +121,7 @@ class UwsgiPMDA(PMDA):
         self.timeout = DEFAULT_TIMEOUT
         self.host = DEFAULT_HOST
         self.port = DEFAULT_PORT
+        self.get_failed = False
         self.read_config()
         self.build_url()
         self.connect_pmcd()
@@ -178,8 +179,10 @@ class UwsgiPMDA(PMDA):
     def refresh_all(self):
         try:
             stats = requests.get(self.url, timeout=self.timeout).json()
+            self.get_failed = False
             #self.log("refresh_all stats:", str(stats))
         except Exception:
+            self.get_failed = True
             return
 
         avg_response_time_msec = 0
@@ -301,5 +304,12 @@ class UwsgiPMDA(PMDA):
             return self.uwsgi_fetch_worker_callback(item, inst)
         return [PM_ERR_PMID, 0]
 
+    def domain_probe(self):
+        """
+        check: return True if server is available and monitorable
+        """
+        self.refresh_all()
+        return self.get_failed
+
 if __name__ == '__main__':
     UwsgiPMDA('uwsgi', 161).run()

src/pmdas/uwsgi/uwsgi.pmcheck

@@ -0,0 +1,40 @@
+#!/bin/sh
+#
+# uWSGI PMDA "plugin" for pmcheck
+#
+
+. $PCP_DIR/etc/pcp.env || exit 1
+. $PCP_SHARE_DIR/lib/checkproc.sh
+
+_do_args "$@"
+
+_check()
+{
+    test -n "$@" && echo "$@" >> $tmp/out
+    [ "$verbose" -gt 0 -a -s $tmp/out ] && cat $tmp/out
+    [ $status -eq 0 ] || exit
+}
+
+iam=uwsgi
+if $lflag
+then
+    [ "$verbose" -gt 0 ] && echo "uWSGI PMDA - metrics from uWSGI servers"
+elif $sflag
+then
+    status=0  # assume active until proven not to be
+    _ctl_pmda state $iam || status=$?
+    PROBE="$PCP_PMDAS_DIR/$iam/pmda$iam.python --probe"
+    PCP_PYTHON_PROBE=1 $PCP_PYTHON_PROG $PROBE || status=2
+    _check "$iam PMDA status: $status"
+elif $aflag
+then
+    _ctl_pmda activate $iam pmda$iam.python || status=1
+elif $dflag
+then
+    _ctl_pmda deactivate $iam || status=1
+else
+    [ $verbose -gt 0 ] && echo "botch sflag=$sflag aflag=$aflag dflag=$dflag show_me=$show_me verbose=$verbose"
+    status=99
+fi
+
+exit

src/python/pcp/pmda.py

@@ -20,6 +20,7 @@
 # pylint: disable=too-many-arguments,consider-using-dict-items,no-member
 
 import os
+import sys
 
 import cpmapi
 import cpmda
@@ -455,6 +456,14 @@ def __init__(self, name, domain, logfile=None, helpfile=None):
     ##
     # general PMDA class methods
 
+    def domain_probe(self):
+        """
+        Probe the domain to see if the PMDA could be activated
+        Used by pmcheck(1) - see man page for meaning of (int)
+        return codes - as part of PMDA specific scripts.
+        """
+        return 99  # unknown, subclasses override to use this.
+
     def domain_write(self):
         """
         Write out the domain.h file (used during installation)
@@ -492,13 +501,18 @@ def pmda_ready(self):
     def run(self):
         """
         All the real work happens herein; we can be called in one of three
-        situations, determined by environment variables.  First couple are
-        during the agent Install process, where the domain.h and namespace
-        files need to be created.  The third case is the real mccoy, where
-        an agent is actually being started by pmcd/dbpmda and makes use of
-        libpcp_pmda to talk PCP protocol.
+        situations, determined by environment variables.  First one is for
+        pmcheck(1), next two are part of the agent Install process (where
+        the domain.h and namespace files need to be created).  The fourth
+        case is the real mccoy, where an agent is actually being started
+        by pmcd/dbpmda and makes use of libpcp_pmda to talk PCP protocol.
         """
-        if 'PCP_PYTHON_DOMAIN' in os.environ:
+        if 'PCP_PYTHON_PROBE' in os.environ:
+            result = self.domain_probe()
+            if isinstance(result, int):
+                sys.exit(int(result))
+            sys.exit(2)
+        elif 'PCP_PYTHON_DOMAIN' in os.environ:
             self.domain_write()
         elif 'PCP_PYTHON_PMNS' in os.environ:
             self.pmns_write(os.environ['PCP_PYTHON_PMNS'])
@@ -568,6 +582,8 @@ def set_notify_change():
 
     @staticmethod
     def set_user(username):
+        if 'PCP_PYTHON_PROBE' in os.environ:
+            return cpmapi.PM_ERR_NOTCONN
         return cpmapi.pmSetProcessIdentity(username)
 
     @staticmethod

src/python/pmda.c

@@ -940,6 +940,9 @@ endContextCallBack(int ctx)
 static inline int
 pmda_generating_pmns(void) { return getenv("PCP_PYTHON_PMNS") != NULL; }
 
+static inline int
+pmda_probing_domain(void) { return getenv("PCP_PYTHON_PROBE") != NULL; }
+
 static inline int
 pmda_generating_domain(void) { return getenv("PCP_PYTHON_DOMAIN") != NULL; }
 
@@ -984,7 +987,7 @@ init_dispatch(PyObject *self, PyObject *args, PyObject *keywords)
     pmdaSetFetchCallBack(&dispatch, fetch_callback);
     pmdaSetEndContextCallBack(&dispatch, endContextCallBack);
 
-    if (!pmda_generating_pmns() && !pmda_generating_domain())
+    if (!pmda_generating_pmns() && !pmda_generating_domain() && !pmda_probing_domain())
 	pmdaOpenLog(&dispatch);
 
     Py_INCREF(Py_None);
@@ -1003,7 +1006,7 @@ connect_pmcd(void)
      * channel setup and complete the connection handshake with
      * pmcd.
      */
-    if (!pmda_generating_pmns() && !pmda_generating_domain()) {
+    if (!pmda_generating_pmns() && !pmda_generating_domain() && !pmda_probing_domain()) {
 	/*
 	 * On success pmdaConnect sets PMDA_EXT_CONNECTED in e_flags ...
 	 * this used in the guard below to stop pmda_dispatch() calling