I am trying to customize OpenWhisk to call a microkernel from the Invoker, rather than Docker. Is there an effort underway currently to add this support, or a development guide covering the changes I would need to make? My current understanding of the code is that this will be a substantial project.
EDIT: To ask the question more pointedly, is there guidance available on how to move away from the concept of containers? Or will I be better off treating a microkernel as an abstracted type of container?
Given OpenWhisk is an Apache project, I believe this question is a perfect fit for the relevant mailing-list. You'll find all the contributors there and it is much more geared for a discussion (which is needed for this topic I think).
Please refer to http://openwhisk.incubator.apache.org/contact.html for more information on how to reach the dev-list.
This PR that integrated Mesos into OpenWhisk may give you an idea of the required changes: https://github.com/apache/incubator-openwhisk/pull/2833
Related
So I moved to a new version of OpenDayLight, im currently using Neon (Sodiums api explorer wasnt working for whatever reason but im trying again today).
I remember older versions of OpenDayLight had a L2Switch but I guess it is no longer supported, I was messing around last night installing flows so 2 hosts on a single switch can communicate. I was doing this through restconf and the md-sal:add-flow or whatever its called and it wasnt very efficient but it did work and I was able to pass traffic.
I may write my own L2 learning code cause ive done it for other controllers and its fairly easy but it seems like that is an outdated approach to getting reachability in an SDN network and im starting to feel like im missing something.
Is there some feature in ODL Sodium/Neon that handles ARP for hosts, or does some type of L2 learning? If not, whats an effective way of installing flows into a network, specifically a Data centre style network with lots of switches. Manually installing flows on one switch was bearable but once my hardware is ready we will have 8 physical OVS devices and installing flows one at a time on each using rest may kill me lmao.
Also, I've been looking around here and through project code for Packet IN handling on the controller, I will need to do this if I do L2 learning and my main goal with ODL is to develop some type of network security application.
If im not mistaken, when an OVS device receives a packet it doesn't know how to forward, it will send it to the controller and the controller will make a decision and install the flow on the switch for future packets. I know how to install the flow, im just a little lost on where to put the code to check the packet and decide what to do with it. In POX it was easy, pretty much just a PacketIn function that you would override, hopefully its something similar with ODL
It also appears Oxygen still has L2Switch, if I have a large topology that I cant manually install flows in, would it be worth it to downgrade to oxygen for L2 functionality?
yeah, I think what you really want is l2switch. that project was always the
quick project to install with OpenDaylight to show some quick SDN functionality
(using mininet), but it was not really maintained in the upstream community
so we removed it. However, there has been talk lately of trying to keep it
alive so at least it can be released and installed in future ODL versions.
would you consider helping in that effort? If so, you could get l2switch working
again the latest master branch (what will be the magnesium release) and possibly
in to one of the final sodium service releases (SR). I don't know all that's
needed to make that happen, but I think it may not be too hard.
taking that one step back would really help the upstream ODL community, but
hopefully give you what you need in the short term with a newer ODL release
as well as some more experience with the overall ODL code-base.
If not, you are on the right track with programming flows with restconf. You
don't have to use md-sal:add-flow necessarily, you can just add the flows
you want in the config store and as long as the openflow node exists, the
openflowplugin should program it down on the ovs switch. Here is a test
case that adds the "punt to controller" flow that you refer to. just drill
down in to the robot framework keywords to get more details. like here is the
PUT request URI and here is the body sent.
I am aware of the process to install WAS 8.5.5.x and 9.0.x versions using IM response file(s) but would like to know best practices and recommendations to perform WAS installation and upgrade on more than one server, to avoid manual errors and reduce time.
I am open to use to Ansible, Puppet or any other orchestration tools as well, but would like to know possible options if we are not allowed to use these tools.
Ultimate goal is to automate most of the setup/upgrade steps, if not all of them since when dealing with bunch of servers.
Thanks
Assuming you are referring to WebSphere Application Server traditional, take a look at the approaches described here, https://www.ibm.com/support/knowledgecenter/SSEQTP_9.0.0/com.ibm.websphere.installation.base.doc/ae/tins_enterprise_install.html, especially if you are working with larger scale deployments.
Consider creating master images and distributing them in a swinging profile-type setup. They make it easier and faster to install and apply updates since you only need to create images once and distribute many times. You have consistency across systems too.
You can then automate with your preferred automation technology.
We use ansible, simple and effectively.
True, you must of course develop a playbook that will be able to do all this.
Im doing research regarding continuous integration tools and there benefits. For my research im looking at the following tools:
GitLab CI
Jenkins
Bamboo
GoCD
TeamCity
Now I wont bother you with all the requirements and benefits. But so far im not finding so many differences between the tools except for these:
Fan-in fan-out support GoCD
Community size, Jenkins and GitLab seem to have most contributors
Costs
Open source or not
Amount of plugins available
I was wondering if some people who have had to choose a continuous integration tool aswell could share there experience and why they chose that tool and if there are certain differences that are worth thinking about before choosing which I didn't cover.
Now im leaning towards GoCD because of fan-in fan-out support and the visualisation of the continuous delivery pipeline does anybody have experience with the support on issues for this tool?
Thanks in regard,
Disclaimer: I was an active contributor to GoCD before previous Fall.
I haven't used GitLab CI so won't talk about that :) Also, I haven't used any of these tools in the past one year.
I think TeamCity is a good CI tool. It integrates very well with IDE if you want to debug some failures. The test reports are brilliant. But I don't think they are that advanced in CD space and in my opinion you need both. But if you are interested only in CI, you might want to give it a look. However, you will miss on some of the good features of GoCD I've mentioned below.
Jenkins has a huge community but Jenkins has its own disadvantages. Many a times one plugin doesn't work due to another plugin for some compatibility issues for instance.
GoCD has Fan-in/Fan-out support which avoids many unnecessary builds saving a lot of build time and resources. The value stream map is intuitive and helps to get a better picture of the build stage from a developer's, QA's or even Deliver Manager's point of view. The pipeline modeling in GoCD is also very good. If you read Jez Humble and David Farley's book on Continuous Delivery, you will see the power behind such a build design.
Now, to your second question:
Now im leaning towards GoCD because of fan-in fan-out support and the
visualisation of the continuous delivery pipeline does anybody have
experience with the support on issues for this tool?
Good to hear that :P I love GoCD. The support is good. If you choose to go the Open Source way, the mailing list is pretty active. You can expect a reply from the GoCD team within a day or two. Of course, your questions have to be genuine and specific. Looking through the forums before posting a question helps :)
You can also choose to buy support for GoCD from ThoughtWorks. They used to offer multiple support tiers, not sure of the current support model. You might face issues only when your DB grows too huge (~5-7 GB) when you might want to go for the proprietary Postgres DB support from ThoughtWorks. I've seen very few users of GoCD with that DB size.
I have a lot of experience with Teamcity and some with Gocd. If you are interested in fan-in/fan-out it's also possible to do the same in Teamcity -- it's called Build Chains.
Also there is a good post about this topic on official blog.
If I could choose I would prefer Teamcity. It's more mature and more feature rich product suitable for use in corporate environment.
I am new to load testing.
So please help in learning gatling and Apache Jmeter for stress testing.
Please help in installing both on Windows and Linux.
How to implement them in my application?
Which one is better for stress testing?
You are asking very generic questions in terms of Stress/Load testing. I think it would be best if you take a look at their documentation then formulate a more specific question.
Installation documentation is best served from the creators of the software.
Implementing these load/stress testing tools into your application isn't really a thing. If you are looking for unit testing (test to utilize in validating your functions/classes/etc work then look at your languages specific go-to libraries - ie. Java is junit/jboss, Nodejs is Karma/Protractor, Python is TestCase/Nose, etc). These tools (jmeter/gatling) are used for stressing your application outside of your build process so they should be treated as end-users (meaning you run the stress testing from remote machines if it is a web service).
Either are best for the right scenario. I think jmeter clusters easier (built-in, where gatling is more manual) but gatling is more programatic and can be manipulated more.
These are opinions and shouldn't be taken as fact or the best so your milage may vary
I strongly doubt that you need them both, if you want a piece of advice in regards which one to choose take a look at Open Source Load Testing Tools: Which One Should You Use? guide.
Once you have clear vision on what tool better suits the needs - you could start ramping up on the selected tool and ask questions in its community communication channels.
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 8 years ago.
Improve this question
I'm facing developing a new web app in the future and I'm wondering how to decide what framework to use. I've settled on Python as my language of choice. But there are still may frameworks to choose from! More generally how do you choose between different similar technologies that are still in the works as the latest round of web frameworks are? I'm curious what your process is for deciding on technologies you've never used.
Recognize that no choice is perfect -- or even very good.
No matter what you choose, someone will have a suggestion that -- they claim -- is better.
No matter what you choose, some part of your tech. stack will fail to live up to your expectations.
The most important thing is "shared nothing" so that the components can be replaced.
After that, the next most important thing is automatically-generated features to reduce or prevent programming.
Look at Django. Lots of automatic admin features make life very pleasant.
There are a number of things you can do:
Download the frameworks and build something similar with them for comparison.
Look for comparisons by other people, but attempt to understand the bias of the reviewer.
Observe the community at work, see what people are building and the issues they run into when using the technology. Forums, blogs, mailing list etc are good places to check out.
Go to conferences and meet like minded developers interested.
You can also take the approach of using stable versions rather than alpha bits. After a while you might move closer the bleeding edge. People associated with the project in question are generally more biased than those approaching from other platforms, be careful who you trust.
Consider the impact of using a bleeding edge framework versus an established one. Sometimes it's important to your customers that you are on one perceived as stable. At other times this doesn't matter. How comfortable are you with fixing the framework itself? Great developers will learn the internals, or at least know enough to keep things moving whilst a bug is sent to the framework mailing list etc.
Consider some general best practices in building abstractions and reusable code on the python platform. You may be able to save yourself some work in moving to another platform. However, don't be a reuse junkie as this can limit the effectiveness of your use of the framework. The 37Signals guys are right when they talk about extracting frameworks from working code rather than building frameworks from scratch.
I know this is an old posting, but I am in a similar situation (again) and I think there are other people who may want to look for different opinions, and hear of (somewhat) successful experiences.
Since baudtack mentioned Python, I will try to answer this along the lines of my experiences using Python. Here is what has been working for me:
determine the scope of your project - outlining what your application is supposed to be able to do without introducing any programming or design notes will clarify your goals greatly
determine how you would like to work with your code, stack and data:
a. what sort of programming paradigm do you want to work with? i.e. object-oriented, functional, etc. do you want to play to your programming style or do you want to follow somebody else's programming style?
b. use semantic web or not? do you want greater control over URIs and their design? (I found web.py great for this by the way - It is my choice to create REST APIs in Python)
c. do you want to be trapped by framework requirements, or do you want a better separation of the application from the web component, i.e. use a framework to utilize your application as a set of modules, for example. My problem with Django was that I ended up not programming Python, but having to learn more Django than I needed to. If that works for you, then that is the way to go.
d. data stores... some sort of SQL vs. non RDBMS (xml databases like eXist-db with full xquery support) vs. OODBMS vs. a combination of the above? how complicated do you need this to be? how much control/separation do you need to have over how data gets stored and recalled in your application?
e. testing: unit tests... thank goodness for python! if your web app has the potential to grow (as they often do), having a sane and coherent testing platform to begin with will help out a lot in the future - I wish I had learned about this earlier on. oh well... better late than never.
f. how much control over the server do you need? hosting considerations? how much control over an Apache instance do you need to have? OS specific needs? I found that using shared hosting providers like Webfaction has been great. I eventually found I needed greater needs for flexibility and bandwidth. In other words, what can you get for your budget? If you have USD50 to spend each month, it may be better to consider a virtual hosting solution like Linode....
Finally, I echo S.Lott's sentiments that no choice for a solution is perfect, and are subject to obsolescence.
Experience trumps hearsay. I've found that prototyping is a huge help. Make a prototype that uses the features you expect to be the most important for various frameworks. This helps route out any features that may not work "as advertised."
In general though, kudos for being willing to look at new technologies.
I have a set of criteria in different categories:
Activity & Documentation
Is there an active user base?
Is there an active development base?
Is the support responsive and information accessible?
Are there user and development guides and reference material?
These are essential, there needs to be traceability of all of these to build confidence in the solution.
Ease of use
Are basic features easy and complex features possible? I typically give a new framework a test drive and try to roll out a set of use cases to see how intuitive the framework is to use.
Is installation intuitive and simple for a local/dev installation and production deployment?
How is it backed up and upgraded?
What is the effort and UX for implementing a "Hello World" type blog post, static page, menu item, and plugin?
How are versions dealt with for the core & plugins?
Example (on the topic of Automated Testing/Continuous Integration solutions)
Several years ago I evaluated several Automated Testing solution. At the time Jenkins and TeamCity were front runners and in the end I chose TeamCity because of the UX, active user & development base and quality of accessible documentation.
Example (CMS for a blog)
This criteria is also why I prefer to use Wordpress over other options. While wordpress has its shortcomings, the user and development base is strong and active which leads to a software architecture with more potential to evolve over time and maintain its relevance and a development community that provides quality plugins and themes to choose from.