Nail/fission-src
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add gevent based Python server to benchmark test cases (#794)

Browse Source 
This add the benchmark test cases, which is using gevent based on wsgi
framework.
This commit is contained in:
xiekeyang
2018-08-01 21:22:03 +08:00
committed by Ta-Ching Chen
parent 9a7ffc2f5b
commit 6123873d2a
7 changed files with 210 additions and 66 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/benchmark/assets/envs/python_bjoern/README
+42
View File
@@ -0,0 +1,42 @@
Fission Specs
=============
This is a set of specifications for a Fission app. This includes functions,
environments, and triggers; we collectively call these things "resources".
How to use these specs
----------------------
These specs are handled with the 'fission spec' command. See 'fission spec --help'.
'fission spec apply' will "apply" all resources specified in this directory to your
cluster. That means it checks what resources exist on your cluster, what resources are
specified in the specs directory, and reconciles the difference by creating, updating or
deleting resources on the cluster.
'fission spec apply' will also package up your source code (or compiled binaries) and
upload the archives to the cluster if needed. It uses 'ArchiveUploadSpec' resources in
this directory to figure out which files to archive.
You can use 'fission spec apply --watch' to watch for file changes and continuously keep
the cluster updated.
You can add YAMLs to this directory by writing them manually, but it's easier to generate
them. Use 'fission function create --spec' to generate a function spec,
'fission environment create --spec' to generate an environment spec, and so on.
You can edit any of the files in this directory, except 'fission-deployment-config.yaml',
which contains a UID that you should never change. To apply your changes simply use
'fission spec apply'.
fission-deployment-config.yaml
------------------------------
fission-deployment-config.yaml contains a UID. This UID is what fission uses to correlate
resources on the cluster to resources in this directory.
All resources created by 'fission spec apply' are annotated with this UID. Resources on
the cluster that are _not_ annotated with this UID are never modified or deleted by
fission.
test/benchmark/assets/envs/python_bjoern/env.yaml
+20
View File
@@ -0,0 +1,20 @@
apiVersion: fission.io/v1
kind: Environment
metadata:
name: python
namespace: default
spec:
version: 2
poolsize: 3
TerminationGracePeriod: 5
resources:
limits:
cpu: 300m
memory: 256Mi
requests:
cpu: 300m
memory: 256Mi
runtime:
image: fission/python-env
imagePullPolicy: IfNotPresent
test/benchmark/assets/envs/python_bjoern/fission-deployment-config.yaml
+7
View File
@@ -0,0 +1,7 @@
# This file is generated by the 'fission spec init' command.
# See the README in this directory for background and usage information.
# Do not edit the UID below: that will break 'fission spec apply'
apiVersion: fission.io/v1
kind: DeploymentConfig
name: envs
uid: a0f67a18-7d1d-4e50-9d9a-dab854e01106
test/benchmark/assets/envs/python_gevent/README
+42
View File
@@ -0,0 +1,42 @@
Fission Specs
=============
This is a set of specifications for a Fission app. This includes functions,
environments, and triggers; we collectively call these things "resources".
How to use these specs
----------------------
These specs are handled with the 'fission spec' command. See 'fission spec --help'.
'fission spec apply' will "apply" all resources specified in this directory to your
cluster. That means it checks what resources exist on your cluster, what resources are
specified in the specs directory, and reconciles the difference by creating, updating or
deleting resources on the cluster.
'fission spec apply' will also package up your source code (or compiled binaries) and
upload the archives to the cluster if needed. It uses 'ArchiveUploadSpec' resources in
this directory to figure out which files to archive.
You can use 'fission spec apply --watch' to watch for file changes and continuously keep
the cluster updated.
You can add YAMLs to this directory by writing them manually, but it's easier to generate
them. Use 'fission function create --spec' to generate a function spec,
'fission environment create --spec' to generate an environment spec, and so on.
You can edit any of the files in this directory, except 'fission-deployment-config.yaml',
which contains a UID that you should never change. To apply your changes simply use
'fission spec apply'.
fission-deployment-config.yaml
------------------------------
fission-deployment-config.yaml contains a UID. This UID is what fission uses to correlate
resources on the cluster to resources in this directory.
All resources created by 'fission spec apply' are annotated with this UID. Resources on
the cluster that are _not_ annotated with this UID are never modified or deleted by
fission.
test/benchmark/assets/envs/python_gevent/env.yaml
+23
View File
@@ -0,0 +1,23 @@
apiVersion: fission.io/v1
kind: Environment
metadata:
name: python
namespace: default
spec:
version: 2
poolsize: 3
TerminationGracePeriod: 5
resources:
limits:
cpu: 300m
memory: 256Mi
requests:
cpu: 300m
memory: 256Mi
runtime:
image: fission/python-env
imagePullPolicy: IfNotPresent
container:
env:
- name: WSGI_FRAMEWORK
value: "GEVENT"
test/benchmark/assets/envs/python_gevent/fission-deployment-config.yaml
+7
View File
@@ -0,0 +1,7 @@
# This file is generated by the 'fission spec init' command.
# See the README in this directory for background and usage information.
# Do not edit the UID below: that will break 'fission spec apply'
apiVersion: fission.io/v1
kind: DeploymentConfig
name: envs
uid: 1fc02a0c-7a51-4534-a86f-a8f397dfd1f2
test/benchmark/tests/concurrency-level/run.sh
+69 -66
View File
@@ -2,92 +2,95 @@
set -euo pipefail
ROOT=$(dirname $0)/../../../..
ROOT=$(readlink -f ./../../../..)
for executorType in poolmgr newdeploy
for framework in python_bjoern python_gevent
do
for concurrency in 100 250 500 750 1000
do
for executorType in poolmgr newdeploy
do
for concurrency in 100 250 500 750 1000
do
testDuration="60"
dirName="concurrency-${concurrency}-executor-${executorType}"
testDuration="60"
dirName="concurrency-${concurrency}-executor-${executorType}"
# remove old data
rm -rf ${dirName}
mkdir ${dirName}
pushd ${dirName}
# remove old data
rm -rf ${dirName}
mkdir ${dirName}
pushd ${dirName}
# run multiple iterations to reduce impact of imbalance of pod distribution.
for iteration in {1..10}
do
# run multiple iterations to reduce impact of imbalance of pod distribution.
for iteration in {1..10}
do
# Create a hello world function in nodejs, test it with an http trigger
echo "Pre-test cleanup"
fission env delete --name python || true
# Create a hello world function in nodejs, test it with an http trigger
echo "Pre-test cleanup"
fission env delete --name python || true
echo "Creating python env"
# Use short grace period time to speed up resource recycle time
# Use high min/max CPU so that K8S will distribute pod in different nodes
fission env create --name python --version 2 --image fission/python-env --period 5 --mincpu 300 --maxcpu 300 --minmemory 256 --maxmemory 256
trap "fission env delete --name python" EXIT
echo "Creating python env"
# Use short grace period time to speed up resource recycle time
# Use high min/max CPU so that K8S will distribute pod in different nodes
fission spec apply --specdir $ROOT/test/benchmark/assets/envs/$framework
trap "fission env delete --name python" EXIT
sleep 30
sleep 30
fn=python-hello-$(date +%s)
fn=python-hello-$(date +%s)
echo "Creating package"
rm -rf pkg.zip pkg/ || true
mkdir pkg
cp ../../../assets/hello.py pkg/hello.py
echo "Creating package"
rm -rf pkg.zip pkg/ || true
mkdir pkg
cp $ROOT/test/benchmark/assets/hello.py pkg/hello.py
zip -jr pkg.zip pkg/
pkgName=$(fission pkg create --env python --deploy pkg.zip | cut -d' ' -f 2 | cut -d"'" -f 2)
zip -jr pkg.zip pkg/
pkgName=$(fission pkg create --env python --deploy pkg.zip | cut -d' ' -f 2 | cut -d"'" -f 2)
echo "Creating function"
fission fn create --name $fn --env python --pkg ${pkgName} --entrypoint "hello.main" --executortype ${executorType} --minscale 3 --maxscale 3
echo "Creating function"
fission fn create --name $fn --env python --pkg ${pkgName} --entrypoint "hello.main" --executortype ${executorType} --minscale 3 --maxscale 3
echo "Creating route"
fission route create --function $fn --url /$fn --method GET
echo "Creating route"
fission route create --function $fn --url /$fn --method GET
echo "Waiting for router to catch up"
sleep 5
echo "Waiting for router to catch up"
sleep 5
fnEndpoint="http://$FISSION_ROUTER/$fn"
js="sample.js"
rawFile="raw-${iteration}.json"
rawUsageReport="raw-usage.txt"
fnEndpoint="http://$FISSION_ROUTER/$fn"
js="sample.js"
rawFile="raw-${iteration}.json"
rawUsageReport="raw-usage.txt"
k6 run \
-e FN_ENDPOINT="${fnEndpoint}" \
--duration "${testDuration}s" \
--rps ${concurrency} \
--vus ${concurrency} \
--no-connection-reuse \
--out json="${rawFile}" \
--summary-trend-stats="avg,min,med,max,p(5),p(10),p(15),p(20),p(25),p(30),p(35),p(40),p(45),p(50),p(55),p(60),p(65),p(70),p(75),p(80),p(85),p(90),p(95),p(100)" \
../${js} >> ${rawUsageReport}
k6 run \
-e FN_ENDPOINT="${fnEndpoint}" \
--duration "${testDuration}s" \
--rps ${concurrency} \
--vus ${concurrency} \
--no-connection-reuse \
--out json="${rawFile}" \
--summary-trend-stats="avg,min,med,max,p(5),p(10),p(15),p(20),p(25),p(30)" \
../${js} >> ${rawUsageReport}
echo "Clean up"
fission fn delete --name ${fn}
fission env delete --name python
fission route list| grep ${fn}| awk '{print $1}'| xargs fission route delete --name
fission pkg delete --name ${pkgName}
rm -rf pkg.zip pkg
echo "Clean up"
fission fn delete --name ${fn}
fission env delete --name python
fission route list| grep ${fn}| awk '{print $1}'| xargs fission route delete --name
fission pkg delete --name ${pkgName}
rm -rf pkg.zip pkg
kubectl -n fission-function get deploy -o name|xargs -I@ bash -c "kubectl -n fission-function delete @" || true
kubectl -n fission-function get pod -o name|xargs -I@ bash -c "kubectl -n fission-function delete @" || true
kubectl -n fission-function get deploy -o name|xargs -I@ bash -c "kubectl -n fission-function delete @" || true
kubectl -n fission-function get pod -o name|xargs -I@ bash -c "kubectl -n fission-function delete @" || true
echo "All done."
done
echo "All done."
done
usageReport="usage.txt"
cat ${rawUsageReport}| grep "http_req_duration"| cut -f2 -d':' > ${usageReport}
usageReport="usage.txt"
cat ${rawUsageReport}| grep "http_req_duration"| cut -f2 -d':' > ${usageReport}
popd
popd
# generate report after iterations are over
outImage="${dirName}.png"
picasso -file ${dirName} -format png -o ${outImage}
# generate report after iterations are over
outImage="${dirName}.png"
picasso -file ${dirName} -format png -o ${outImage}
done
done
done
done