Hi Andres, I reply inline.
Regards, Manuel On Tue, 2017-07-11 at 16:00 +0200, [email protected] wrote: > Manuel, > > Thank you for your proposal it has been very helpful, i have > already > deployed the Openstack environment and it looks good, it passed the > health check made by fuel, i also deployed a VM and connected to it > successfully and have ODL up and running. > > I am having a problem to make the functest because i am doing it > inside a workstation that does not have access to the Openstack > management network, it is located inside the public network. I am > not > sure which is the best way to proceed: > > A) if I give the workstation an IP address inside the management > network will it work? or maybe IP tables are configured to reject > that. iptables should be ok, you need to make sure that the route is correct. Let me understand this, you have three physical machines: 1 - Runs OpenStack 2 - Runs FUEL 3 - Workstation where you want to run functest ?? > B) Do the functest from my Fuel master (it is a PC that only runs > Fuel). > > I read the posts in your blog that you sent Rosa and found them > extremely interesting, it has been very educative. I have some > questions regarding that scenario that you built (If you have no > problem answering them): > - What OS and application do you use for deploying the SF? do you use > OVS? We have a qcow2 image that we use as SF: http://artifacts.opnfv.org/sfc /images/sfc_nsh_danube.qcow2 That one has a tool which reads nsh packets and pushes them back to the SFF. Try to understand the different tests we have in the SFC-OPNFV repo and you will see how we do that. For example this one: https://github.com/opnfv/sfc/blob/master/sfc/tests/functest/sfc_one_cha in_two_service_functions.py Which is explained here: https://wiki.opnfv.org/display/sfc/Functest+SFC-ODL+-+Test+2 > - Once you deploy a SFC does tacker pushes the configuration to ODL > via REST? or do you configure the SFF and SF manually on ODL? tacker pushes all the config, you don't have to mess with ODL Rest API > - Do you have any more tutorials or videos similar? We did one last year: https://www.youtube.com/watch?v=QknJMX83q1k&t=345s And maybe these ones help you understand things better: https://www.youtube.com/watch?v=b2xjpllQ7d8 https://www.youtube.com/watch?v=KKKqvXFRJZ0 > - I have read that OVS+NSH implementation that you use is developed > by > OPNFV and is not mainstream in OVS. Is it possible to use that > OVS+NSH > on a physical device running openWRT? The patch which provides NSH to OVS is not only used in OPNFV. Hopefully, it will get upstreamed by OVS2.8 (~September). I am not sure if the patch is supported by that architecture. Please ask his maintainer: [email protected] > > I ask all this questions because we are interested in setting up a > similar scenario! > > Thank you in advance for your response. > > Best regards, > > Quoting "Manuel Buil" <[email protected]>: > > > > > Hi Andres, > > > > Please find my answers below. > > > > Try to run the environment which is supported and after that you > > can > > add stuff like ceilometer. Having a base that works should be your > > target right now, later you can add stuff to that base. > > > > Regards, > > Manuel > > > > On Mon, 2017-07-10 at 13:51 +0200, > > [email protected] wrote: > > > > > > Hello Manuel, > > > > > > Thank you for your quick response! you are right that scenario is > > > not > > > supported sorry for that! I am trying to set up a NFV > > > environment > > > (using SFC) with the ability to monitor cloud resources using > > > Openstack APIs, and according to what i have been able to read > > > ceilometer provides that information! > > > > > > Ok since no SFC chains are supposed to be declared i am guessing > > > i > > > am > > > close to deploying the scenario correctly, but i am having > > > these > > > issues (let me know if you how to address them): > > > > > > - Once I create a VM I am not being able to access them (ping > > > fails). > > > > If you list the VM with 'nova list', does it say it is active? Does > > it > > list an ip? Check that it receives an IP lease from the server with > > nova console-log > > > > > > > > > > - When I restart a compute node it will take an IP address from > > > my > > > public network and after that I cannot ssh into the node. > > > > Strange... but why do you reboot a compute? Are you trying to test > > some > > HA behaviour? > > > > > > > > > > > I have the following doubts regarding the deployment options > > > being > > > set > > > up in fuel: > > > > > > - When creating the environment in fuel what networking option > > > should > > > i choose: Neutron with ML2 plugin & Neutron with tunneling > > > segmentation or OpenDaylight with tunneling segmentation? > > > > If you want to try SFC, you should use OpenDaylight > > > > > > > > - Right now i am choosing the following role distribution: > > > controller-ceph, controller-tacker, controller-ODL, compute- > > > ceph, > > > compute-ceph. Is this appropriate? > > > > Start with a simpler env. with one controller and one compute. The > > controller should act as cotroller, OpenDaylight controller and > > Tacker > > VNF Manager. Ceph should work but better simpligy things and use > > 'Cinder LVM' > > > > > > > > - I am using the option that states: Assign public network to > > > all > > > nodes (I read in the guide that this should be checked). But i > > > think > > > this is causing the communication issue to the nodes. > > > > I have that option checked too but I never reboot computes ;) > > > > > > > > - I install Open vSwitch with the checkbox that says install NSH. > > > Is > > > this correct? > > > > Yes > > > > > > > > - When marking the ODL Plugin I only check the box that says: > > > SFC > > > features with NetVirt classifier. What about use ODL to manage > > > L3 > > > traffic? should I mark it. > > > > It is not needed for sfc. That one allows you to connect several > > openstack deployments among them using L3VPN. > > > > > > > > - Do I need to install any other features in ODL (i.e l2switch) > > > in > > > order to communicate with my VMs or do I need to declare some > > > SFC > > > chains? > > > > Nothing else. When the deployment succeeds, you should be able to > > run > > tests. To run them, you should follow this guide (note that we just > > realized that Danube 2.0 is throwing errors... try Danube 1.0 > > instead): > > > > https://wiki.opnfv.org/display/sfc/OPNFV-SFC+Functest+test+cases > > > > > > > > > > > Thank you in advance for your help, i tried to write as clear as > > > i > > > could but english is not my native tongue > > > > You write very clearly! > > > > > > > > > > > Quoting "Manuel Buil" <[email protected]>: > > > > > > > > > > > > > > > Hi Andres, > > > > > > > > Unfortunately, that scenario is not supported in Danube. These > > > > are > > > > the > > > > ones supported: > > > > > > > > https://wiki.opnfv.org/display/SWREL/Danube+Scenario+Status > > > > > > > > What statistics do you need from SFC? Maybe you can collect > > > > them in > > > > another way. > > > > > > > > When the sfc scenarios are successfully deployed, no VMs or SFC > > > > chains > > > > exist. > > > > > > > > Regards, > > > > Manuel > > > > > > > > > > > > On Fri, 2017-07-07 at 10:58 +0200, > > > > [email protected] wrote: > > > > > > > > > > > > > > > Hello, > > > > > > > > > > My name is Andrés Sánchez, i am working on deploying the > > > > > OPNFV > > > > > HA > > > > > scenario on a laboratory on my university. I have been trying > > > > > to > > > > > deploy the scenario manually but have been encountering some > > > > > problems. > > > > > My actual environment consists of: > > > > > > > > > > 5 Nodes: 2 CPUs, 8GB RAM. > > > > > 1 Fuel master: 2 CPUs, 8GB RAM. > > > > > 1 workstation: 2CPUs, 4GB RAM. > > > > > > > > > > I am trying to deploy the > > > > > "ha_odl-l2_sfc_heat_ceilometer_scenario.yaml" on this lab, i > > > > > have > > > > > previously deployed other Openstack scenarios on this lab so > > > > > the > > > > > networking configuration is properly configured: on each > > > > > Openstack > > > > > node port i assigned untagged traffic to PXE-admin, vlan 200 > > > > > to > > > > > public > > > > > network, and defaults to storage,management and private > > > > > (102,101,103). > > > > > PXE-admin network is 192.168.100.0/24 and public network is > > > > > 172.16.10.0/24 with gateways being .1 respectively. > > > > > > > > > > I already set up the fuel master with the Danube ISO > > > > > "opnfv-danube.1.0.iso". I previously correctly deployed an > > > > > openstack > > > > > cluster manually installing ODL and tacker plugins, but then > > > > > i > > > > > found > > > > > that did not appear to be correctly configured and neither > > > > > ODL > > > > > (I > > > > > created a couple of VMs and could not connect to them). When > > > > > creating > > > > > the environment which option is supposed to be selected: > > > > > Neutron > > > > > with > > > > > ML2 plugin & Neutron with tunneling segmentation or > > > > > OpenDaylight > > > > > with > > > > > tunneling segmentation. Another thing i found is that fuel > > > > > did > > > > > not > > > > > provide the option for adding Telemetry-Mongo DB in node > > > > > assignment > > > > > (i > > > > > am requiring to have Telemetry because i need to check the > > > > > meters > > > > > it > > > > > provides). I did check the box for ?Assign public network to > > > > > all > > > > > nodes? > > > > > > > > > > I am trying to deploy now from my workstation but i have > > > > > doubts > > > > > regarding the way the DHA-DEA files are supposed to be > > > > > written, i > > > > > am > > > > > interested in: > > > > > A) Using the actual fuel master as fuel: > > > > > > > > > > sudo bash ./deploy.sh -b > > > > > file:///home/lab232/fuel/deploy/config > > > > > -f > > > > > -l > > > > > devel-pipeline -p lab232 -s > > > > > ha_odl-l2_sfc_heat_ceilometer_scenario.yaml -i > > > > > file:///home/lab232/opnfv-danube.1.0.iso -e > > > > > > > > > > B) Create a VM inside my workstation to host Fuel master and > > > > > use > > > > > all > > > > > the other nodes as Openstack nodes. > > > > > > > > > > sudo bash ./deploy.sh -b > > > > > file:///home/lab232/fuel/deploy/config > > > > > -F > > > > > -l > > > > > devel-pipeline -p lab232 -s > > > > > ha_odl-l2_sfc_heat_ceilometer_scenario.yaml -i > > > > > file:///home/lab232/opnfv-danube.1.0.iso -e > > > > > > > > > > For both scenarios i have encountered several issues an > > > > > errors, i > > > > > am > > > > > attaching the configuration files i use. Could you please > > > > > provide > > > > > some > > > > > guidance in the correct way to write these files. > > > > > > > > > > When i try to deploy scenario A i am finding ipmi adapter > > > > > errors: > > > > > Exception: Address lookup for None failed > > > > > Could not open socket! > > > > > Error: Unable to establish IPMI v2 / RMCP+ session > > > > > > > > > > When i try to deploy scenario B i am finding: "Exception: > > > > > Device > > > > > "pxebr" does not exist." I am guessing i need to declare a > > > > > bridge > > > > > called pxebr on my host machine, please confirm. > > > > > > > > > > I am attaching my Dea/Dha files! > > > > > > > > > > One final question: when the environment is correctly > > > > > deployed > > > > > is > > > > > openstack supposed to have some VMs declared? is ODL supposed > > > > > to > > > > > have > > > > > any SFC paths or something declared? > > > > > > > > > > Thank you in advance for your help! > > > > > > > > > > Best regards, > > > > > > > > > > _______________________________________________ > > > > > opnfv-tech-discuss mailing list > > > > > [email protected] > > > > > https://lists.opnfv.org/mailman/listinfo/opnfv-tech-discuss > > > > > > > > > > > > > > > > > > _______________________________________________ opnfv-tech-discuss mailing list [email protected] https://lists.opnfv.org/mailman/listinfo/opnfv-tech-discuss
