From 282369b6fd58a78e6a7c91f21b331363d4ed0fb3 Mon Sep 17 00:00:00 2001 From: Umar Farooq Date: Thu, 13 Jul 2017 12:20:56 +0200 Subject: Add Collectd as a Monitor Type A plugin for collectd is added to use collectd on compute as a monitor type. Monitor files are updated accordingly. The inspector now listens on all interfaces instead of only localhost to enable it to communicate with compute node. JIRA: DOCTOR-86 JIRA: DOCTOR-101 Change-Id: Idc834d428152e4687020eff7d8db36a652b1bf86 Signed-off-by: Umar Farooq --- tests/run.sh | 31 ++++++++++++++++--------------- 1 file changed, 16 insertions(+), 15 deletions(-) (limited to 'tests/run.sh') diff --git a/tests/run.sh b/tests/run.sh index abdad58c..69f18b20 100755 --- a/tests/run.sh +++ b/tests/run.sh @@ -212,17 +212,6 @@ create_alarm() { done } -start_monitor() { - pgrep -f "python monitor.py" && return 0 - sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \ - > monitor.log 2>&1 & -} - -stop_monitor() { - pgrep -f "python monitor.py" || return 0 - sudo kill $(pgrep -f "python monitor.py") -} - start_consumer() { pgrep -f "python consumer.py" && return 0 python consumer.py "$CONSUMER_PORT" > consumer.log 2>&1 & @@ -294,8 +283,12 @@ inject_failure() { echo "disabling network of compute host [$COMPUTE_HOST] for 3 mins..." cat > disable_network.sh << 'END_TXT' #!/bin/bash -x -dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $NF}') sleep 1 +if [ -n "@INTERFACE_NAME@" ]; then + dev=@INTERFACE_NAME@ +else + dev=$(sudo ip a | awk '/ @COMPUTE_IP@\//{print $NF}') +fi sudo ip link set $dev down echo "doctor set link down at" $(date "+%s.%N") sleep 180 @@ -303,6 +296,7 @@ sudo ip link set $dev up sleep 1 END_TXT sed -i -e "s/@COMPUTE_IP@/$COMPUTE_IP/" disable_network.sh + sed -i -e "s/@INTERFACE_NAME@/$INTERFACE_NAME/" disable_network.sh chmod +x disable_network.sh scp $ssh_opts_cpu disable_network.sh "$COMPUTE_USER@$COMPUTE_IP:" ssh $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP" 'nohup ./disable_network.sh > disable_network.log 2>&1 &' @@ -327,8 +321,11 @@ calculate_notification_time() { wait_consumer 60 #keep 'at' as the last keyword just before the value, and #use regex to get value instead of the fixed column + if [ ! -f monitor.log ]; then + scp $ssh_opts_cpu "$COMPUTE_USER@$COMPUTE_IP:monitor.log" . + fi detected=$(grep "doctor monitor detected at" monitor.log |\ - sed -e "s/^.* at //") + sed -e "s/^.* at //" | tail -1) notified=$(grep "doctor consumer notified at" consumer.log |\ sed -e "s/^.* at //" | tail -1) @@ -431,11 +428,11 @@ run_profiler() { cleanup() { set +e echo "cleanup..." - stop_monitor stop_inspector stop_consumer unset_forced_down_hosts + stop_monitor collect_logs vms=$(openstack $as_doctor_user server list) @@ -467,6 +464,7 @@ cleanup() { cleanup_installer cleanup_inspector + cleanup_monitor # NOTE: Temporal log printer. for f in $(find . -name '*.log') @@ -511,6 +509,9 @@ setup_python_packages source $TOP_DIR/functions-common source $TOP_DIR/lib/installer source $TOP_DIR/lib/inspector +source $TOP_DIR/lib/monitor + +rm -f *.log setup_installer @@ -540,8 +541,8 @@ echo "injecting host failure..." inject_failure check_host_status "(DOWN|UNKNOWN)" 60 -calculate_notification_time unset_forced_down_hosts +calculate_notification_time collect_logs run_profiler -- cgit 1.2.3-korg