Minutes of Bottlenecks Discussion on July 20, 2017
* Date and Time
* Thursday at
0100-0200<http://www.timeanddate.com/worldclock/fixedtime.html?hour=13&min=0&sec=0>
UTC, July 20, 2017
* Thursday at 0900-1000 Beijing, July 20, 2017
* Wednesday at 1800-1900 PDT, July 20, 2017
* Minutes
* Yang Yu
* Participants (9 peoples):
* Ace Lee
* Jack Chen
* Jing Lu
* Kubi
* Manuel Rebellon
* Rex
* Ross Brattain
Agenda:
* Discussion with Yardstick about the scaling test
* Stress Testing Discuss & Release Discuss
* Action Items Review
Discussion & Action Item
1. Discussion with Yardstick about the scaling test
* https://etherpad.opnfv.org/p/yardstick_release_e
* Scale-out test
* Scale-up test
* Ross briefly introduced the scaling test cases
* The test cases are similar to the netperf and ping tests in
Bottlenecks repo
* Ross asks about how Bottlenecks passes parameters to Yardstick
* There is a separated config file located in Bottlenecks repo and
modified test case yaml file located in Yardstick's. Then Bottlenecks pass the
stack number to Yardstick in each testing iteration
*
https://gerrit.opnfv.org/gerrit/gitweb?p=bottlenecks.git;a=blob;f=testsuites/posca/testcase_cfg/posca_factor_ping.yaml;h=ed1e3475321668d50f4dcdd658d1f00f579e1f77;hb=HEAD
*
https://gerrit.opnfv.org/gerrit/gitweb?p=yardstick.git;a=blob;f=samples/ping_bottlenecks.yaml;h=01977a1dee1f0c9318ce272609d4ad9a253d7625;hb=HEAD
* Some modifications of the testing codes are also needed to
accommodate the changes in Yardstick
* Bottlenecks chooses a more stable way to call Yardstick by using
docker library to call docker exec command making Yardstick run certain test
case
* Action on Ross to provide an example yaml for Scaling and discuss with
Bottleneck. Then Bottlenecks implement the testing part
* Ross asks about how Bottlenecks process the results
* Bottlenecks get the testing result through the result file produced by
yardstick filtering the KPIs that Bottlenecks would like to collect. Then plot
the KPIs vs each iteration in Kibana and the monitored time series results in
Grafana.
* Gabriel is also introduce the monitoring tools that recently be merged
in Bottlenecks named Prometheus that monitoring the Host status
* Ross asks about the stop criteria for the testing iterations
* Currently Bottlenecks examines the increasement/decreasement for the
monitoring results.
* Once the increasement/decreasement between 2 consecutive iterations is
below the preset threshold, e.g., 2.5.%. Then we stop the test.
2. Stress Testing Discuss & Release Discuss
* Long duration POD for stress testing
* Report from last TSC meeting
* Questions raised about the choice of SUT and stability criteria
* Jose will organize a discuss the Testperf meeting focusing on
the plan
* Need to go to TSC next week
* Cross project stress is under consideration
* Running VSperf while StorPerf is excuted
* Bottlenecks act as load manager while monitoring/analyzing
* Intern Projects
* Have contacted the intern for CPU limit
* An interview has been held and recruit the Intern for CPU limit
* Action on Gabriel to confirm communication media
* Euphrates Planning updated
* https://wiki.opnfv.org/display/bottlenecks/Bottlenecks+Release+Plan
* Testing Framework Refactoring
* Monitoring module has been installed in Bottlenecks
* Prometheus has been supporting in Bottlencks regarding
installation, collectd and node integration.
* Working on installation Grafana to gether with Prometheus
* Prometheus is a powerful tool for monitoring test process and
alert when the KPI pass certain threshold. Gabriel show the dashboard and
results for Prometheus and Grafana.
* Manual asked Prometheus is open source software
* https://prometheus.io/docs/introduction/overview/
* Initiate Container Testing
* This plan has been assigned low priority
* Prakash should feedback about Kola deployment/testing
* K8S, OpenRetriever, ...
* Initiate Tuning Testing
* This plan has been assigned medium priority
* Action item on Gabriel to reorganize the priority
* WIKI page should be elaborated
* https://wiki.opnfv.org/pages/viewpage.action?pageId=12386549
* Ceph or CPU tuning is under consideration now
* Need investigation work
* Action item on Gabriel
3. Project Engagement
* More contributors are welcome to Bottlenecks
* Cooperation with other projects
* Integration with upstream
* Non-active committer contact
* Has sent out emails to confirm the involvement plan of committers for
E release
4. Action Item Review
* https://wiki.opnfv.org/display/bottlenecks/Release+Plan
Euphrates:
* How to use Quota - Prakash
* Reliability project in Openstack - Prakash
* WIKI page restructured for closely updating project details @done
(17-05-11 09:20)
* Testing framework
* General env pre commond for all test cases? Maybe different test
cases call diffrent functions?
* Requirements investigation of DPDK and SR-IOV in Bottlenecks
* Which categories we should put them in?
* What the structure of their report?
* Is there any existing implementation we could use?
* Do we need to separate the env pre and test executing operations?
* Test report
* (Automatically) Test result report to MongoDB @done
* Landing page (on-going, in plan)
* Stress testing
* Implementing the rest stress test cases discussed during Danube
* Collect stress test case
* Request/maintaining long duration pod
* Work with VSPERF & StorPerf
* Initiate Tuning Test
* Ceph Tuning ( installer deployment tuning?)
* Comparison results
* Initiate Container Testing
* K8S, OpenRetriever, unicore
* Work with installers team for Kola deployment testing
* Prakash to check OpenStack for Kola testing
* Testing Framework Refactoring
* Overall framework design
* Abandon docker-compose usage and adaptively testing sourcing OPNFV
credentials.
* Determine requirements & Initiate the overall framework design
* Unified entrance/top logic for system limit/tuning test
* Initiate top logic design & investigation
* Separated monitoring module
* Work with StorPerf & VSPERF adding monitoring function
* Monitoring using yardstick? Barometer?
* Stand-alone & offline
* More Contributors in Bottlenecks
* Intern projects
* Integrate upstream
* More contributors in project
_______________________________________________
opnfv-tech-discuss mailing list
opnfv-tech-discuss@lists.opnfv.org
https://lists.opnfv.org/mailman/listinfo/opnfv-tech-discuss
