Add Rocestat PMDA for collecting and analyzing RoCE device metrics

build/rpm/pcp.spec.in

@@ -1570,6 +1570,20 @@ This package contains the PCP Performance Metrics Domain Agent (PMDA) for
 collecting metrics from simple network checks.
 # end pcp-pmda-netcheck
 
+#
+# pcp-pmda-rocestat
+#
+%package pmda-rocestat
+License: GPL-2.0-or-later
+Summary: Performance Co-Pilot (PCP) metrics for nVidia RoCE devices
+URL: https://pcp.io
+Requires: pcp = @package_version@ pcp-libs = @package_version@
+Requires: python3-pcp
+%description pmda-rocestat
+This package contains the PCP Performance Metrics Domain Agent (PMDA) for
+collecting statistics for nVidia RDMA over Converged Ethernet (RoCE) devices.
+# end pcp-pmda-rocestat
+
 #
 # pcp-pmda-openvswitch
 #
@@ -2242,6 +2256,7 @@ basic_manifest | keep '(etc/pcp|pmdas)/postgresql(/|$)' >pcp-pmda-postgresql-fil
 basic_manifest | keep '(etc/pcp|pmdas)/rabbitmq(/|$)' >pcp-pmda-rabbitmq-files
 basic_manifest | keep '(etc/pcp|pmdas)/redis(/|$)' >pcp-pmda-redis-files
 basic_manifest | keep '(etc/pcp|pmdas)/resctrl(/|$)|sys-fs-resctrl' >pcp-pmda-resctrl-files
+basic_manifest | keep '(etc/pcp|pmdas)/rocestat(/|$)' >pcp-pmda-rocestat-files
 basic_manifest | keep '(etc/pcp|pmdas)/roomtemp(/|$)' >pcp-pmda-roomtemp-files
 basic_manifest | keep '(etc/pcp|pmdas)/rsyslog(/|$)' >pcp-pmda-rsyslog-files
 basic_manifest | keep '(etc/pcp|pmdas)/samba(/|$)' >pcp-pmda-samba-files
@@ -2279,7 +2294,7 @@ for pmda_package in \
     nutcracker nvidia \
     openmetrics openvswitch oracle \
     pdns perfevent podman postfix postgresql \
-    rabbitmq redis resctrl roomtemp rsyslog \
+    rabbitmq redis resctrl rocestat roomtemp rsyslog \
     samba sendmail shping slurm smart snmp \
     sockets statsd summary systemd \
     unbound uwsgi \
@@ -2696,6 +2711,9 @@ done
 %preun pmda-rabbitmq
 %{pmda_remove "$1" "rabbitmq"}
 
+%preun pmda-rocestat
+%{pmda_remove "$1" "rocestat"}
+
 %preun pmda-uwsgi
 %{pmda_remove "$1" "uwsgi"}
 %endif
@@ -3020,6 +3038,8 @@ fi
 
 %files pmda-rabbitmq -f pcp-pmda-rabbitmq-files.rpm
 
+%files pmda-rocestat -f pcp-pmda-rocestat-files.rpm
+
 %files pmda-uwsgi -f pcp-pmda-uwsgi-files.rpm
 
 %files export-pcp2graphite -f pcp-export-pcp2graphite-files.rpm

build/rpm/redhat.spec

@@ -1729,6 +1729,25 @@ collecting metrics from simple network checks.
 # end pcp-pmda-netcheck
 %endif
 
+#
+# pcp-pmda-rocestat
+#
+%package pmda-rocestat
+License: GPL-2.0-or-later
+Summary: Performance Co-Pilot (PCP) metrics for nVidia RoCE devices
+URL: https://pcp.io
+Requires: pcp = %{version}-%{release} pcp-libs = %{version}-%{release}
+%if !%{disable_python3}
+Requires: python3-pcp
+%else
+Requires: %{__python2}-pcp
+%endif
+%description pmda-rocestat
+This package contains the PCP Performance Metrics Domain Agent (PMDA) for
+collecting statistics for nVidia RDMA over Converged Ethernet (RoCE) devices.
+# end pcp-pmda-rocestat
+%endif
+
 %if !%{disable_mongodb}
 #
 # pcp-pmda-mongodb
@@ -2449,6 +2468,7 @@ basic_manifest | keep '(etc/pcp|pmdas)/postgresql(/|$)' >pcp-pmda-postgresql-fil
 basic_manifest | keep '(etc/pcp|pmdas)/rabbitmq(/|$)' >pcp-pmda-rabbitmq-files
 basic_manifest | keep '(etc/pcp|pmdas)/redis(/|$)' >pcp-pmda-redis-files
 basic_manifest | keep '(etc/pcp|pmdas)/resctrl(/|$)|sys-fs-resctrl' >pcp-pmda-resctrl-files
+basic_manifest | keep '(etc/pcp|pmdas)/rocestat(/|$)' >pcp-pmda-rocestat-files
 basic_manifest | keep '(etc/pcp|pmdas)/roomtemp(/|$)' >pcp-pmda-roomtemp-files
 basic_manifest | keep '(etc/pcp|pmdas)/rpm(/|$)' >pcp-pmda-rpm-files
 basic_manifest | keep '(etc/pcp|pmdas)/rsyslog(/|$)' >pcp-pmda-rsyslog-files
@@ -2487,7 +2507,7 @@ for pmda_package in \
     nutcracker nvidia \
     openmetrics openvswitch oracle \
     pdns perfevent podman postfix postgresql \
-    rabbitmq redis resctrl roomtemp rpm rsyslog \
+    rabbitmq redis resctrl rocestat roomtemp rpm rsyslog \
     samba sendmail shping slurm smart snmp \
     sockets statsd summary systemd \
     unbound uwsgi \
@@ -2865,6 +2885,9 @@ exit 0
 %preun pmda-netcheck
 %{pmda_remove "$1" "netcheck"}
 
+%preun pmda-rocestat
+%{pmda_remove "$1" "rocestat"}
+
 %endif
 
 %preun pmda-apache
@@ -3209,6 +3232,8 @@ fi
 
 %files pmda-rabbitmq -f pcp-pmda-rabbitmq-files.rpm
 
+%files pmda-rocestat -f pcp-pmda-rocestat-files.rpm
+
 %files pmda-uwsgi -f pcp-pmda-uwsgi-files.rpm
 
 %files export-pcp2graphite -f pcp-export-pcp2graphite-files.rpm

qa/1993

@@ -0,0 +1,87 @@
+#!/bin/sh
+# PCP QA Test No. 1993
+# Testing PCP Rocestat PMDA - install, remove and values.
+#
+# Copyright (c) 2025 Oracle and/or its affiliates.
+# DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+#
+# 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
+# Free Software Foundation; either version 2 of the License, or (at your
+# option) any later version.
+#
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+# or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+# for more details.
+#
+
+seq=`basename $0`
+echo "QA output created by $seq"
+
+. ./common.python
+
+pmda_path="$PCP_PMDAS_DIR/rocestat"
+pmda_script="$pmda_path/pmdarocestat.python"
+test_script="rocestat/rocestat_test.python"
+
+[ -d $pmda_path ] || _notrun "ROCESTAT PMDA is not installed"
+
+$python -c "from pcp import pmda" >/dev/null 2>&1
+[ $? -eq 0 ] || _notrun "python pcp pmda module not installed"
+
+if [ ! -d /sys/class/infiniband ]; then
+    # Check if Infiniband kernel module is loaded
+    lsmod | grep -q '^ib_core' || _notrun "IB kernel modules are not loaded"
+    _notrun "No RoCE devices detected"
+fi
+
+status=1	# failure is the default!
+trap "_cleanup_pmda rocestat; exit \$status" 0 1 2 3 15
+
+pmdarocestat_filter()
+{
+    sed \
+        -e "s/Ran [0-9]* tests in [0-9]*\.[0-9][0-9]*s/Ran X tests in YYYs/" \
+        -e "s/FAILED (failures=[0-9]*)/FAILED (failures=X)/" \
+        -e "/^$/d"
+}
+
+pmdarocestat_remove()
+{
+    cd $pmda_path
+    echo
+    echo "=== Removing ROCESTAT agent ==="
+    $sudo ./Remove >$tmp.out 2>&1
+}
+
+pmdarocestat_install()
+{
+    cd $pmda_path
+    $sudo ./Remove >/dev/null 2>&1
+
+    echo
+    echo "=== Installing ROCESTAT agent ==="
+    $sudo ./Install </dev/null >$tmp.out 2>&1
+    cd $here
+}
+
+run_rocestat_test()
+{
+    echo
+    echo "=== Running ROCESTAT test script ==="
+    $sudo $python $test_script 2>&1 | LC_COLLATE=POSIX sort | pmdarocestat_filter
+}
+
+# Real QA test starts here
+_prepare_pmda rocestat
+
+pmdarocestat_install
+
+run_rocestat_test
+
+pmdarocestat_remove
+
+# Success, all done
+status=0
+exit

qa/1993.out

@@ -0,0 +1,21 @@
+QA output created by 1993
+
+=== Installing ROCESTAT agent ===
+
+=== Running ROCESTAT test script ===
+----------------------------------------------------------------------
+Fetching Rocestat PMDA metrics
+OK
+Ran X tests in YYYs
+test_hw_link_metrics (__main__.RocestatTests) ... ok
+test_hw_mcast_metrics (__main__.RocestatTests) ... ok
+test_hw_metrics (__main__.RocestatTests) ... ok
+test_hw_rcv_metrics (__main__.RocestatTests) ... ok
+test_hw_req_metrics (__main__.RocestatTests) ... ok
+test_hw_resp_metrics (__main__.RocestatTests) ... ok
+test_hw_rnr_metrics (__main__.RocestatTests) ... ok
+test_hw_ucast_metrics (__main__.RocestatTests) ... ok
+test_hw_xmit_metrics (__main__.RocestatTests) ... ok
+test_nic_metrics (__main__.RocestatTests) ... ok
+
+=== Removing ROCESTAT agent ===

qa/GNUmakefile

@@ -17,7 +17,7 @@ TESTS	= $(shell sed -n -e '/^[0-9][0-9]*:retired/d' -e '/^[0-9][0-9]*:reserved/d
 SUBDIRS = src pmdas cisco gluster pconf sadist collectl nfsclient named \
 	  archives badarchives views qt linux unbound cifs gpfs lustre ganglia \
 	  java mmv postfix perl json slurm tmparch sheet smart admin hacluster \
-	  sockets denki gfs2 farm pdudata
+	  sockets denki gfs2 farm pdudata rocestat
 
 ifeq "$(PMDA_PERFEVENT)" "true"
 SUBDIRS += perfevent

qa/GNUmakefile.install

@@ -21,7 +21,7 @@ ifeq "$(PMDA_PERFEVENT)" "true"
 SUBDIRS += perfevent
 endif
 ifeq "$(HAVE_PYTHON)" "true"
-SUBDIRS += secure mic haproxy lio openmetrics
+SUBDIRS += secure mic haproxy lio openmetrics rocestat
 endif
 
 # Before installing the pcpqa policy module we need a PCP build with

qa/group

@@ -260,6 +260,7 @@ pmda.postgresql
 pmda.proc
 pmda.redis
 pmda.resctrl
+pmda.rocestat
 pmda.root
 #pmda.rpm		# note this group has been retired
 pmda.rsyslog
@@ -2228,4 +2229,5 @@ pmcd.pdu
 1990 pcp buddyinfo python local
 1991 pcp netstat python local
 1992 pmda.uwsgi local
+1993 pmda.rocestat local python
 4751 libpcp threads valgrind local pcp helgrind

qa/rocestat/GNUmakefile

@@ -0,0 +1,22 @@
+#!gmake
+
+TOPDIR = ../..
+include $(TOPDIR)/src/include/builddefs
+
+TESTDIR = $(PCP_VAR_DIR)/testsuite/rocestat
+MYFILES = rocestat.json
+PYMODULES = rocestat_test.python
+LDIRT = rocestat_test.py rocestat_test.pyc rocestat_test.pyo
+
+default default_pcp setup: rocestat_test.py
+
+install install_pcp:
+	$(INSTALL) -m 755 -d $(TESTDIR)
+	$(INSTALL) -m 644 -f $(MYFILES) $(TESTDIR)/$(MYFILES)
+	$(INSTALL) -m 755 -f $(PYMODULES) $(TESTDIR)/$(PYMODULES)
+	$(INSTALL) -m 644 -f GNUmakefile.install $(TESTDIR)/GNUmakefile
+
+include $(BUILDRULES)
+
+%.py : %.python
+	$(LN_S) $< $@

qa/rocestat/GNUmakefile.install

@@ -0,0 +1,20 @@
+#!gmake
+
+ifdef PCP_CONF
+include $(PCP_CONF)
+else
+include $(PCP_DIR)/etc/pcp.conf
+endif
+PATH	= $(shell . $(PCP_DIR)/etc/pcp.env; echo $$PATH)
+include $(PCP_INC_DIR)/builddefs
+
+TESTDIR = $(PCP_VAR_DIR)/testsuite/rocestat
+
+default default_pcp setup: rocestat_test.py
+
+install install_pcp:
+
+include $(BUILDRULES)
+
+%.py : %.python
+	$(LN_S) $< $@

qa/rocestat/README

@@ -0,0 +1,29 @@
+Pre-Requisites
+--------------
+In order to perform the tests, we'll make some assumptions:
+
+- The `pcp` Python module is installed and accessible.
+- The `rocestat` PMDA has been installed successfully.
+- The system has RoCE devices available and the required kernel modules (`ib_core`) are loaded.
+- You are able to run the test as the root user (required for PMDA installation and removal).
+
+What the Test Does
+------------------
+The test validates the functionality of the `rocestat` PMDA by performing the following steps:
+
+1. Installs the `rocestat` PMDA.
+2. Fetches metrics defined in `metrics.list` and validates their values against the expected values in `rocestat.json`.
+3. Runs unit tests for each metric cluster (e.g., `hw`, `hw_xmit`, `hw_rcv`) to ensure the PMDA is reporting correct values.
+4. Removes the `rocestat` PMDA after the tests are complete.
+
+If any of the above checks fail, an exception will be raised, and the test will terminate.
+
+How to Test
+-----------
+1. Make sure the pre-requisites are in place.
+2. Run the rocestat_test.python script
+
+Issues
+------
+- None known at this time
+

qa/rocestat/metrics.list

@@ -0,0 +1,49 @@
+rocestat.lane.rx_pause
+rocestat.lane.rx_bytes
+rocestat.lane.tx_bytes
+rocestat.hw.link.link_error_recovery
+rocestat.hw.link.link_downed
+rocestat.hw.link.local_link_integrity_errors
+rocestat.hw.rnr.rnr_nak_retry_err
+rocestat.hw.resp.resp_local_length_error
+rocestat.hw.resp.resp_cqe_error
+rocestat.hw.resp.resp_cqe_flush_error
+rocestat.hw.resp.resp_remote_access_errors
+rocestat.hw.req.req_remote_invalid_request
+rocestat.hw.req.req_cqe_error
+rocestat.hw.req.req_cqe_flush_error
+rocestat.hw.req.duplicate_request
+rocestat.hw.req.rx_read_requests
+rocestat.hw.req.rx_atomic_requests
+rocestat.hw.req.req_remote_access_errors
+rocestat.hw.req.rx_write_requests
+rocestat.hw.mcast.multicast_rcv_packets
+rocestat.hw.mcast.multicast_xmit_packets
+rocestat.hw.ucast.unicast_rcv_packets
+rocestat.hw.ucast.unicast_xmit_packets
+rocestat.hw.rcv.port_rcv_errors
+rocestat.hw.rcv.port_rcv_remote_physical_errors
+rocestat.hw.rcv.port_rcv_packets
+rocestat.hw.rcv.port_rcv_data
+rocestat.hw.rcv.port_rcv_constraint_errors
+rocestat.hw.rcv.port_rcv_switch_relay_errors
+rocestat.hw.xmit.port_xmit_data
+rocestat.hw.xmit.port_xmit_constraint_errors
+rocestat.hw.xmit.port_xmit_wait
+rocestat.hw.xmit.port_xmit_packets
+rocestat.hw.xmit.port_xmit_discards
+rocestat.hw.roce_slow_restart_trans
+rocestat.hw.roce_slow_restart_cnps
+rocestat.hw.roce_slow_restart
+rocestat.hw.roce_adp_retrans_to
+rocestat.hw.clear_counters
+rocestat.hw.local_ack_timeout_err
+rocestat.hw.lifespan
+rocestat.hw.implied_nak_seq_err
+rocestat.hw.packet_seq_err
+rocestat.hw.roce_adp_retrans
+rocestat.hw.out_of_buffer
+rocestat.hw.out_of_sequence
+rocestat.hw.VL15_dropped
+rocestat.hw.excessive_buffer_overrun_errors
+rocestat.hw.symbol_error