Re: [openstack-dev] [oslo][messaging][zmq] Discussion on zmq driver design issues
On Tue, Mar 10, 2015 at 8:14 PM, ozamiatin ozamia...@mirantis.com wrote: Hi Li Ma, Thank you very much for your reply It's good to hear you have a living deployment with zmq driver. Is there a big divergence between your production and upstream versions of the driver? Besides [1] and [2] fixes for redis we have [5] and [6] critical multi-backend related issues for using the driver in real-world deployment. Actually there's no such a big divergence between our driver and upstream version. We didn't refactor it much, but just fixed all the bugs that we met before and implemented socket reuse mechanism to greatly improve its performance. For some bugs available, especially cinder multi-backend and neutron multi-worker, we hacked cinder and neutron to get rid of these bugs. I discussed with our cinder developer several times about these problems you mentioned above. Due to the current architecture, it is really difficult to fix it in zeromq driver. However, it is very easy to deal with it in cinder. We have patches on hand, but the implementation is a little tricky that the upstream may not accept it. :-( No worry, I'll find out it soon. By the way, we are discussing about fanout performance and message persistance. I don't have codes available, but I've got some ideas to implement it. The only functionality for large-scale deployment that lacks in the current upstream codebase is socket pool scheduling (to provide lifecycle management, including recycle and reuse zeromq sockets). It was done several months ago and we are willing to contribute. I plan to propose a blueprint in the next release. Pool, recycle and reuse sounds good for performance. Yes, actually our implementation is a little ugly and there's no unit test available. Right now, I'm trying to refactor it and hopefully I'll submit a spec soon. We also need a refactoring of the driver to reduce redundant entities or reconsider them (like ZmqClient or InternalContext) and to reduce code replications (like with topics). There is also some topics management needed. Clear code == less bugs == easy understand == easy contribute. We need a discussion (with related spec and UMLs) about what the driver architecture should be (I'm in progress with that). +1, cannot agree with you more. 3. ZeroMQ integration I've been working on the integration of ZeroMQ and DevStack for a while and actually it is working right now. I updated the deployment guide [3]. That's true it works! :) I think it is the time to bring a non-voting gate for ZeroMQ and we can make the functional tests work. You can turn it with 'check experimental'. It is broken now. I'll figure it out soon. 5. ZeroMQ discussion Here I'd like to say sorry for this driver. Due to spare time and timezone, I'm not available for IRC or other meeting or discussions. But if it is possible, should we create a subgroup for ZeroMQ and schedule meetings for it? If we can schedule in advance or at a fixed date time, I'm in. That's great idea +1 for zmq subgroup and meetings I'll open another thread to discuss this topic. Subfolder is actually what I mean (python package like '_drivers') it should stay in oslo.messaging. Separate package like oslo.messaging.zeromq is overkill. As Doug proposed we can do consistently to AMQP-driver. I suggest you go for it right now. It is really important for further development. If I submit new codes based upon the current code structure, it will greatly affect this work in the future. Best regards, -- Li Ma (Nick) Email: skywalker.n...@gmail.com __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [ceilometer] Pipeline for notifications does not seem to work
I am just curious have you restarted ceilometer services after pipeline.yaml has been changed? Igor Degtiarov Software Engineer Mirantis Inc www.mirantis.com On Sat, Mar 21, 2015 at 1:21 PM, Tim Bell tim.b...@cern.ch wrote: No errors in the notification logs. Should this work with the default ceilometer.conf file or do I need to enable anything ? I’ve also tried using arithmetic. When I have a meter like “cpu” for the source, this fires the expression evaluation without problems. However, I can’t find a good way of doing the appropriate calculations using the number of cores. Sample calculation is below expr: $(cpu)*0.98+$(vcpus)*10.0 I have tried $(cpu.resource_metdata.vcpus) and $(cpu.resource_metdata.cpu_number) also. Any suggestions on an alternative approach that could work ? Any suggestions for the variable name to get at the number of cores when I’m evaluating an expression fired by the cpu time ? Tim From: gordon chung [mailto:g...@live.ca] Sent: 20 March 2015 20:55 To: OpenStack Development Mailing List not for usage questions Subject: Re: [openstack-dev] [ceilometer] Pipeline for notifications does not seem to work i can confirm it works for me as well... are there any noticeable errors in the ceilometer-agent-notifications log? the snippet below looks sane to me though. cheers, gord From: idegtia...@mirantis.com Date: Fri, 20 Mar 2015 18:35:56 +0200 To: openstack-dev@lists.openstack.org Subject: Re: [openstack-dev] [ceilometer] Pipeline for notifications does not seem to work Hi Tim I've check your case on my devstack. And I've received new hs06 meter in my meter list. So something wrong with your local env. Cheers, Igor D. Igor Degtiarov Software Engineer Mirantis Inc www.mirantis.com On Fri, Mar 20, 2015 at 5:40 PM, Tim Bell tim.b...@cern.ch wrote: I’m running Juno with ceilometer and trying to produce a new meter which is based on vcpus * F (where F is a constant that is different for each hypervisor). When I create a VM, I get a new sample for vcpus. However, it does not appear to fire the transformer. The same approach using “cpu” works OK but this one is polling on a regular interval rather than a one off notification when the VM is created. Any suggestions or alternative approaches for how to get a sample based the number of cores scaled by a fixed constant? Tim In my pipeline.yaml sources, - name: vcpu_source interval: 180 meters: - vcpus sinks: - hs06_sink In my transformers, I have - name: hs06_sink transformers: - name: unit_conversion parameters: target: name: hs06 unit: HS06 type: gauge scale: 47.0 publishers: - notifier:// __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova] how to handle vendor-specific API microversions?
On 03/21/2015 01:21 AM, Chris Friesen wrote: Hi, I've recently been playing around a bit with API microversions and I noticed something that may be problematic. The way microversions are handled, there is a monotonically increasing MAX_API_VERSION value in nova/api/openstack/api_version_request.py. When you want to make a change you bump the minor version number and it's yours. End-users can set the microversion number in the request to indicate what they support, and all will be well. The issue is that it doesn't allow for OpenStack providers to add their own private microversion(s) to the API. They can't just bump the microversion internally because that will conflict with the next microversion bump upstream (which could cause problems when they upgrade). In terms of how to deal with this, it would be relatively simple to just bump the major microversion number at the beginning of each new release. However, that would make it difficult to backport bugfixes/features that use new microversions since they might overlap with private microversions. I think a better solution might be to expand the existing microversion API to include a third digit which could be considered a private microversion, and provide a way to check the third digit separate from the other two. That way providers would have a way to add custom features in a backwards-compatible way without worrying about colliding with upstream code. I would vote that we not make this pleasant or easy for vendors who are wanting to add a feature to the API. As a person who uses several clouds daily, I can tell you that a vendor chosing to do that is VERY mean to users, and provides absolutely no value to anyone, other than allowing someone to make a divergent differentiated fork. Just don't do it. Seriously. It makes life very difficult for people trying to consume these things. The API is not the place for divergence. __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] Gerrit downtime on 2015-03-21
cor...@inaugust.com (James E. Blair) writes: Hi, Gerrit will be unavailable for a few hours starting at 1500 UTC on Saturday, March 21. Gerrit is up and running on the new server. Actual downtime was about 1 hour from 1500 to 1600. Please let us know either here or on Freenode in #openstack-infra if you notice any problems. -Jim __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [oslo][cinder][nova][neutron] going forward to oslo-config-generator ...
Hi, One of the issues that we had in Nova was that when we moved to oslo libraries configuration options support by the libraries were no longer present in the generated configuration file. Is this something that is already supported or planned (sorry for being a little ignorant here). In neutron things may be a little more challenging as there are many different plugins and with the decomposition things may have additional challenges. The configuration binding is done via the external decomposed code and not in the neutron code base. So it is not clear how that code may be parsed to generate the sample configuration. Thanks Gary On 3/21/15, 12:01 AM, Jay S. Bryant jsbry...@electronicjungle.net wrote: All, Let me start with the TLDR; Cinder, Nova and Neutron have lots of configuration options that need to be processed by oslo-config-generator to create the project.conf.sample file. There are a couple of different ways this could be done. I have one proposal out, which has raised concerns, there is a second approach that could be taken which I am proposing below. Please read on if you have a strong opinion on the precedent we will try to set in Cinder. :-) We discussed in the oslo meeting a couple of weeks ago a plan for how Cinder was going to blaze a trail to the new oslo-config-generator. The result of that discussion and work is here: [1] It needs some more work but has the bare bones pieces there to move to using oslo-config-generator. With the change I have written extensive hacking checks that ensure that any lists that are registered with register_opts() are included in the base cinder/opts.py file that is then a single entry point that pulls all of the options together to generate the cinder.conf.sample file. This has raised concern, however, that whenever a developer adds a new list of configuration options, they are going to have to know to go back to cinder/opts.py and add their module and option list there. The hacking check should catch this before code is submitted, but we are possibly setting ourselves up for cases where the patch will fail in the gate because updates are not made in all the correct places and because pep8 isn't run before the patch is pushed. It is important to note, that this will not happen every time a configuration option is changed or added, as was the case with the old check-uptodate.sh script. Only when a new list of configuration options is added which is a much less likely occurrence. To avoid this happening at all it was proposed by the Cinder team that we use the code I wrote for the hacking checks to dynamically go through the files and create cinder/opts.py whenever 'tox -egenconfig' is run. Doing this makes me uncomfortable as it is not consistent with anything else I am familiar with in OpenStack and is not consistent with what other projects are doing to handle this problem. In discussion with Doug Hellman, the approach also seemed to cause him concern. So, I don't believe that is the right solution. An alternative that may be a better solution was proposed by Doug: We could even further reduce the occurrence of such issues by moving the list_opts() function down into each driver and have an entry point for oslo.config.opts in setup.cfg for each of the drivers. As with the currently proposed solution, the developer doesn't have to edit a top level file for a new configuration option. This solution adds that the developer doesn't have to edit a top level file to add a new configuration item list to their driver. With this approach the change would happen in the driver's list_opts() function, rather than in cinder/opts.py . The only time that setup.cfg would needed to edited is when a new package is added or when a new driver is added. This would reduce some of the already minimal burden on the developer. We, however, would need to agree upon some method for aggregating together the options lists on a per package (i.e. cinder.scheduler, cinder.api) level. This approach, however, also has the advantage of providing a better indication in the sample config file of where the options are coming from. That is an improvement over what I have currently proposed. Does Doug's proposal sound more agreeable to everyone? It is important to note that the fact that some manual intervention is required to 'plumb' in the new configuration options was done by design. There is a little more work required to make options available to oslo-config-generator but the ability to use different namespaces, different sample configs, etc were added with the new generator. These additional capabilities were requested by other projects. So, moving to this design does have the potential for more long-term gain. Thanks for taking the time to consider this! Jay [1] https://review.openstack.org/#/c/165431/ __ OpenStack Development Mailing List (not for usage questions) Unsubscribe:
Re: [openstack-dev] [oslo][cinder][nova][neutron] going forward to oslo-config-generator ...
Jay, That sound reasonable. We will need to document in a guide for driver developers what to do when new option is added deprecated in conf file for a driver. Expect that nothing extra will need to be done beyond what we are doing now when new functionality added/deprecated from scheduler/default driver and perculates into drivers a release later. I can also comment directly on the patch if it make sense. Thanks, Arkady -Original Message- From: Jay S. Bryant [mailto:jsbry...@electronicjungle.net] Sent: Friday, March 20, 2015 5:02 PM To: OpenStack Development Mailing List (not for usage questions) Subject: [openstack-dev] [oslo][cinder][nova][neutron] going forward to oslo-config-generator ... All, Let me start with the TLDR; Cinder, Nova and Neutron have lots of configuration options that need to be processed by oslo-config-generator to create the .conf.sample file. There are a couple of different ways this could be done. I have one proposal out, which has raised concerns, there is a second approach that could be taken which I am proposing below. Please read on if you have a strong opinion on the precedent we will try to set in Cinder. :-) We discussed in the oslo meeting a couple of weeks ago a plan for how Cinder was going to blaze a trail to the new oslo-config-generator. The result of that discussion and work is here: [1] It needs some more work but has the bare bones pieces there to move to using oslo-config-generator. With the change I have written extensive hacking checks that ensure that any lists that are registered with register_opts() are included in the base cinder/opts.py file that is then a single entry point that pulls all of the options together to generate the cinder.conf.sample file. This has raised concern, however, that whenever a developer adds a new list of configuration options, they are going to have to know to go back to cinder/opts.py and add their module and option list there. The hacking check should catch this before code is submitted, but we are possibly setting ourselves up for cases where the patch will fail in the gate because updates are not made in all the correct places and because pep8 isn't run before the patch is pushed. It is important to note, that this will not happen every time a configuration option is changed or added, as was the case with the old check-uptodate.sh script. Only when a new list of configuration options is added which is a much less likely occurrence. To avoid this happening at all it was proposed by the Cinder team that we use the code I wrote for the hacking checks to dynamically go through the files and create cinder/opts.py whenever 'tox -egenconfig' is run. Doing this makes me uncomfortable as it is not consistent with anything else I am familiar with in OpenStack and is not consistent with what other projects are doing to handle this problem. In discussion with Doug Hellman, the approach also seemed to cause him concern. So, I don't believe that is the right solution. An alternative that may be a better solution was proposed by Doug: We could even further reduce the occurrence of such issues by moving the list_opts() function down into each driver and have an entry point for oslo.config.opts in setup.cfg for each of the drivers. As with the currently proposed solution, the developer doesn't have to edit a top level file for a new configuration option. This solution adds that the developer doesn't have to edit a top level file to add a new configuration item list to their driver. With this approach the change would happen in the driver's list_opts() function, rather than in cinder/opts.py . The only time that setup.cfg would needed to edited is when a new package is added or when a new driver is added. This would reduce some of the already minimal burden on the developer. We, however, would need to agree upon some method for aggregating together the options lists on a per package (i.e. cinder.scheduler, cinder.api) level. This approach, however, also has the advantage of providing a better indication in the sample config file of where the options are coming from. That is an improvement over what I have currently proposed. Does Doug's proposal sound more agreeable to everyone? It is important to note that the fact that some manual intervention is required to 'plumb' in the new configuration options was done by design. There is a little more work required to make options available to oslo-config-generator but the ability to use different namespaces, different sample configs, etc were added with the new generator. These additional capabilities were requested by other projects. So, moving to this design does have the potential for more long-term gain. Thanks for taking the time to consider this! Jay [1] https://review.openstack.org/#/c/165431/ __ OpenStack Development Mailing List (not for usage
Re: [openstack-dev] [ceilometer] Pipeline for notifications does not seem to work
No errors in the notification logs. Should this work with the default ceilometer.conf file or do I need to enable anything ? I've also tried using arithmetic. When I have a meter like cpu for the source, this fires the expression evaluation without problems. However, I can't find a good way of doing the appropriate calculations using the number of cores. Sample calculation is below expr: $(cpu)*0.98+$(vcpus)*10.0 I have tried $(cpu.resource_metdata.vcpus) and $(cpu.resource_metdata.cpu_number) also. Any suggestions on an alternative approach that could work ? Any suggestions for the variable name to get at the number of cores when I'm evaluating an expression fired by the cpu time ? Tim From: gordon chung [mailto:g...@live.ca] Sent: 20 March 2015 20:55 To: OpenStack Development Mailing List not for usage questions Subject: Re: [openstack-dev] [ceilometer] Pipeline for notifications does not seem to work i can confirm it works for me as well... are there any noticeable errors in the ceilometer-agent-notifications log? the snippet below looks sane to me though. cheers, gord From: idegtia...@mirantis.commailto:idegtia...@mirantis.com Date: Fri, 20 Mar 2015 18:35:56 +0200 To: openstack-dev@lists.openstack.orgmailto:openstack-dev@lists.openstack.org Subject: Re: [openstack-dev] [ceilometer] Pipeline for notifications does not seem to work Hi Tim I've check your case on my devstack. And I've received new hs06 meter in my meter list. So something wrong with your local env. Cheers, Igor D. Igor Degtiarov Software Engineer Mirantis Inc www.mirantis.comhttp://www.mirantis.com On Fri, Mar 20, 2015 at 5:40 PM, Tim Bell tim.b...@cern.chmailto:tim.b...@cern.ch wrote: I'm running Juno with ceilometer and trying to produce a new meter which is based on vcpus * F (where F is a constant that is different for each hypervisor). When I create a VM, I get a new sample for vcpus. However, it does not appear to fire the transformer. The same approach using cpu works OK but this one is polling on a regular interval rather than a one off notification when the VM is created. Any suggestions or alternative approaches for how to get a sample based the number of cores scaled by a fixed constant? Tim In my pipeline.yaml sources, - name: vcpu_source interval: 180 meters: - vcpus sinks: - hs06_sink In my transformers, I have - name: hs06_sink transformers: - name: unit_conversion parameters: target: name: hs06 unit: HS06 type: gauge scale: 47.0 publishers: - notifier:// __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribemailto:openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribemailto:openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova] how to handle vendor-specific API microversions?
On Sat, Mar 21, 2015 at 8:31 AM, Monty Taylor mord...@inaugust.com wrote: On 03/21/2015 01:21 AM, Chris Friesen wrote: Hi, I've recently been playing around a bit with API microversions and I noticed something that may be problematic. The way microversions are handled, there is a monotonically increasing MAX_API_VERSION value in nova/api/openstack/api_version_request.py. When you want to make a change you bump the minor version number and it's yours. End-users can set the microversion number in the request to indicate what they support, and all will be well. The issue is that it doesn't allow for OpenStack providers to add their own private microversion(s) to the API. They can't just bump the microversion internally because that will conflict with the next microversion bump upstream (which could cause problems when they upgrade). In terms of how to deal with this, it would be relatively simple to just bump the major microversion number at the beginning of each new release. However, that would make it difficult to backport bugfixes/features that use new microversions since they might overlap with private microversions. I think a better solution might be to expand the existing microversion API to include a third digit which could be considered a private microversion, and provide a way to check the third digit separate from the other two. That way providers would have a way to add custom features in a backwards-compatible way without worrying about colliding with upstream code. I would vote that we not make this pleasant or easy for vendors who are wanting to add a feature to the API. As a person who uses several clouds daily, I can tell you that a vendor chosing to do that is VERY mean to users, and provides absolutely no value to anyone, other than allowing someone to make a divergent differentiated fork. Just don't do it. Seriously. It makes life very difficult for people trying to consume these things. The API is not the place for divergence. In fact we have made vendorization of the API hard on purpose, see the microversion spec for details: https://review.openstack.org/#/c/127127 To quote Jay Pipes from that review: *-1 for vendor flag* *Recommend getting rid of the vendor: specification entirely. The point of standardizing our APIs is to make them standard, not to allow vendorization. API extensions were an idea designed (in part) to allow vendorization. And we've seen how that works out.* *Let's take a hard stand and say this is the OpenStack Compute API and be done with it. If RAX or HP Cloud or Frobnozzle Cloud wants to have a separate but different API, then they should call it something else, because it's not the OpenStack Compute API* __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev