placement/playbooks/perfload.yaml
Chris Dent 910b466c50 perfload with written allocations
One of the needs we've discussed for perfload is making sure
it is measuring when some inventory has been used.

Here, we change the perload job so that it creates the 1000 providers,
measures getting allocation_candidates and then, in a loop of 99, gets
a limited set of candidates, writes the first one back as an allocation
for a random consumer, project and user. At each iteration it measures
again.

This will make the log file a lot longer, but that's not a significant
issue: the numbers that matter will either be near the top or near the
end. If they are weird, looking in the middle will be informative. We
can tweak it.

This, as usual, is one of many ways to accomplish gathering some data.
Other options might include parallelizing the writes, but in this case
we are trying to see the impact of code on a single request, not on
concurrency.

At some point we will want to add nested and sharing into this mix.

Change-Id: I74b64a25f2be8fbbd01b3a3b438bba68de04b269
2019-06-07 14:27:04 +00:00

154 lines
7.2 KiB
YAML

- hosts: all
tasks:
# Borrowed from devstack tasks but we want it early.
- name: Ensure {{ ansible_user_dir }}/logs exists
become: true
file:
path: "{{ ansible_user_dir }}/logs"
state: directory
owner: "{{ ansible_user }}"
- name: start placement
args:
chdir: "{{ ansible_user_dir }}/src/opendev.org/openstack/placement"
shell:
executable: /bin/bash
cmd: |
set -x
# TODO(cdent): Presumably ansible can do this, perhaps with 'package'.
# create database
sudo debconf-set-selections <<MYSQL_PRESEED
mysql-server mysql-server/root_password password secret
mysql-server mysql-server/root_password_again password secret
mysql-server mysql-server/start_on_boot boolean true
MYSQL_PRESEED
sudo apt-get install -y mysql-server mysql-client libmysqlclient-dev jq
sudo mysql -uroot -psecret -e "DROP DATABASE IF EXISTS placement;"
sudo mysql -uroot -psecret -e "CREATE DATABASE placement CHARACTER SET utf8;"
sudo mysql -uroot -psecret -e "GRANT ALL PRIVILEGES ON placement.* TO 'root'@'%' identified by 'secret';"
python -m virtualenv -p python3 .placement
. .placement/bin/activate
pip install . PyMySQL uwsgi
# set config via environment
export OS_PLACEMENT_DATABASE__CONNECTION=mysql+pymysql://root:secret@127.0.0.1/placement?charset=utf8
export OS_PLACEMENT_DATABASE__MAX_POOL_SIZE=25
export OS_PLACEMENT_DATABASE__MAX_OVERFLOW=100
export OS_PLACEMENT_DATABASE__SYNC_ON_STARTUP=True
# Increase our chances of allocating to different providers.
export OS_PLACEMENT_PLACEMENT__RANDOMIZE_ALLOCATION_CANDIDATES=True
export OS_DEFAULT__DEBUG=True
export OS_API__AUTH_STRATEGY=noauth2
uwsgi --http :8000 --wsgi-file .placement/bin/placement-api --daemonize {{ ansible_user_dir }}/logs/placement-api.log --processes 5 --threads 25
- name: placement performance
args:
chdir: "{{ ansible_user_dir }}/src/opendev.org/openstack/placement"
shell:
executable: /bin/bash
# TODO(cdent): Change this task to a role?
cmd: |
set -x
# Do some performance related information gathering for placement.
EXPLANATION="
This output combines output from placeload with timing information
gathered via curl. The placeload output is the current maximum
microversion of placement followed by an encoded representation of
what it has done. Lowercase 'r', 'i', 'a', and 't' indicate successful
creation of a resource provider and setting inventory, aggregates, and
traits on that resource provider.
If there are upper case versions of any of those letters, a failure
happened for a single request. The letter will be followed by the
HTTP status code and the resource provider uuid. These can be used
to find the relevant entry in logs/screen-placement-api.txt.gz.
Note that placeload does not exit with an error code when this
happens. It merely reports and moves on. Under correct circumstances
the right output is a long string of 4000 characters containing
'r', 'i', 'a', 't' in random order (because async).
After that are three aggregate uuids, timing information for the
placeload run, and then timing information for two identical curl
requests for allocation candidates.
If no timed requests are present it means that the expected number
of resource providers were not created. At this time, only resource
providers are counted, not whether they have the correct inventory,
aggregates, or traits.
"
# This aggregate uuid is a static value in placeload.
AGGREGATE="14a5c8a3-5a99-4e8f-88be-00d85fcb1c17"
TRAIT="HW_CPU_X86_AVX2"
PLACEMENT_QUERY="resources=VCPU:1,DISK_GB:10,MEMORY_MB:256&member_of=${AGGREGATE}&required=${TRAIT}"
PLACEMENT_URL="http://localhost:8000"
LOG=placement-perf.txt
LOG_DEST={{ ansible_user_dir }}/logs
COUNT=1000
trap "sudo cp -p $LOG $LOG_DEST" EXIT
function time_candidates {
(
echo "##### TIMING GET /allocation_candidates?${PLACEMENT_QUERY} twice"
time curl -s -H 'x-auth-token: admin' -H 'openstack-api-version: placement latest' "${PLACEMENT_URL}/allocation_candidates?${PLACEMENT_QUERY}" > /dev/null
time curl -s -H 'x-auth-token: admin' -H 'openstack-api-version: placement latest' "${PLACEMENT_URL}/allocation_candidates?${PLACEMENT_QUERY}" > /dev/null
) 2>&1 | tee -a $LOG
}
function write_allocation {
# Take the first allocation request and send it back as a well-formed allocation
curl -s -H 'x-auth-token: admin' -H 'openstack-api-version: placement latest' "${PLACEMENT_URL}/allocation_candidates?${PLACEMENT_QUERY}&limit=5" \
| jq --arg proj $(uuidgen) --arg user $(uuidgen) '.allocation_requests[0] + {consumer_generation: null, project_id: $proj, user_id: $user}' \
| curl -s -H 'x-auth-token: admin' -H 'content-type: application/json' -H 'openstack-api-version: placement latest' \
-X PUT -d @- "${PLACEMENT_URL}/allocations/$(uuidgen)"
}
function load_candidates {
time_candidates
for iter in {1..99}; do
echo "##### Writing allocation ${iter}" | tee -a $LOG
write_allocation
time_candidates
done
}
function check_placement {
local rp_count
local code
code=0
python -m virtualenv -p python3 .placeload
. .placeload/bin/activate
# install placeload
pip install 'placeload==0.3.0'
set +x
# load with placeload
(
echo "$EXPLANATION"
# preheat the aggregates to avoid https://bugs.launchpad.net/nova/+bug/1804453
placeload $PLACEMENT_URL 10
echo "##### TIMING placeload creating $COUNT resource providers with inventory, aggregates and traits."
time placeload $PLACEMENT_URL $COUNT
) 2>&1 | tee -a $LOG
rp_count=$(curl -H 'x-auth-token: admin' ${PLACEMENT_URL}/resource_providers |json_pp|grep -c '"name"')
# Skip curl and note if we failed to create the required number of rps
if [[ $rp_count -ge $COUNT ]]; then
load_candidates
else
(
echo "Unable to create expected number of resource providers. Expected: ${COUNT}, Got: $rp_count"
echo "See job-output.txt.gz and logs/screen-placement-api.txt.gz for additional detail."
) | tee -a $LOG
code=1
fi
set -x
deactivate
exit $code
}
check_placement