Don't spam postci with deployment time loop
We have a lot of resources in the overcloud Heat stack now, which is causing a lot of noise in the logs. This patch does a couple of things to clean that up: -Shuts off tracing while we're processing deployment times. A couple of log messages should suffice to let people know what is going on at that time. -Limits the number of resources we send to graphite to the top 50. There's no need to be logging times for resources that only take a few seconds, and sending the full list to graphite causes the UI to be really slow loading the list of data points. Change-Id: I07c648761d81350d611257769c1d33803a8f7553
This commit is contained in:
parent
57d6b4b12a
commit
a41025db15
|
@ -216,12 +216,20 @@ function postci(){
|
|||
# post metrics
|
||||
scp $SSH_OPTIONS root@${SEED_IP}:${METRICS_DATA_FILE} /tmp/seed-metrics
|
||||
cat /tmp/seed-metrics >> ${METRICS_DATA_FILE}
|
||||
# This spams the postci output with largely uninteresting trace output
|
||||
set +x
|
||||
echo -n 'Recording Heat deployment times...'
|
||||
# We're only interested in the resources that take the most time
|
||||
long_times=/tmp/long-deploy-times.log
|
||||
head -n 50 $WORKSPACE/logs/undercloud/var/log/heat-deploy-times.log > $long_times
|
||||
while read line; do
|
||||
# $line should look like "ResourceName 123.0", so concatenating all
|
||||
# of this together we should end up with a call that looks like:
|
||||
# record_metric tripleo.overcloud.ha.resources.ResourceName 123.0
|
||||
record_metric tripleo.overcloud.${TOCI_JOBTYPE}.resources.${line}
|
||||
done <$WORKSPACE/logs/undercloud/var/log/heat-deploy-times.log
|
||||
done <$long_times
|
||||
echo 'Finished'
|
||||
set -x
|
||||
metrics_to_graphite "23.253.94.71" #Dan's temp graphite server
|
||||
if [ -z "${LEAVE_RUNNING:-}" ] && [ -n "${HOST_IP:-}" ] ; then
|
||||
destroy_vms &> $WORKSPACE/logs/destroy_vms.log
|
||||
|
|
|
@ -48,6 +48,9 @@ function stop_metric {
|
|||
}
|
||||
|
||||
function metrics_to_graphite {
|
||||
# With a lot of metrics, the trace output gets pretty spammy
|
||||
set +x
|
||||
echo "Sending metrics to graphite"
|
||||
local SERVER=$1
|
||||
local PORT=${2:-2003} # default port for graphite data
|
||||
|
||||
|
@ -60,10 +63,12 @@ function metrics_to_graphite {
|
|||
METRIC_NAME=$(echo $X | cut -d ":" -f 1)
|
||||
METRIC_VAL=$(echo $X | cut -d ":" -f 2)
|
||||
DTS=$(echo $X | cut -d ":" -f 3)
|
||||
echo "Sending $METRIC_NAME to Graphite"
|
||||
echo "$METRIC_NAME $METRIC_VAL $DTS" | nc ${SERVER} ${PORT}
|
||||
done
|
||||
set -e
|
||||
# reset the existing data file and start times
|
||||
echo "" > METRICS_START_TIMES
|
||||
echo "" > METRICS_DATA_FILE
|
||||
set -x
|
||||
}
|
||||
|
|
Loading…
Reference in New Issue