Kin icon

KIN DB 2004 Project - KIN Startup Procedures

Project hosted at: Logo


These are basic instructions to get a basic Kin system running. Here we suppose you already have compiled sources, created directories and sample database, and installed them. If not, please refer to the basic installation instructions before proceeding here.


Starting up Kin cluster

Please read Starting up instructions to configure Kin if you never run it before. Here we describe how to start the whole cluster, provided you configured it well (all the node's IPs and ports are specified into kin.cfg file). Order loading managers is not critical, but as we are working with alpha version, it is better to first load Query Manager, then a Backup/Storage Manager, then an Item Manager, and finally, a Client Manager:

  cd /usr/local/kin
  bin/kinode -mQB
Don't desperate this will take some 10 seconds till the QM has becomed the master QM, and it is safe to load the BM. Wait some 5-10 seconds more to let the BM and QM to become idle and completely initialized, and then:
  bin/kinode -mI
This will init and load objects (after QM decides this -unique- IM has to manage them all). You can have a look at the log files to have an idea of what is happening (as we are testing, a debug level of 4-5 is ok, and a filter with bit1 set, like in number 2). After a few seconds, IM has loaded all the objects (demo database has a very small dataset), so you may proceed to load the last manager:
  bin/kinode -mC
The former loads almost instantly, and provides the client interface. Web Client Interface is still on the first development stage, but it accepts http queries, and has a few features already implemented. Just try something like opening your favourite browser and connecting to (replace by your server address, even the IP address is ok). A page like this appears:
sysinfo page
By the way, if your browser has the ability to display an icon for a web page, this is also supported, just copy your favourite icon to /usr/local/kin/data/htdocs/favicon.ico
Any other page you request will redirect you to the login page, that is not yet finished, so it is normal that you see an error page (light red backgrounded).
Out of the web world, you may also get some reports on systems status just sending a HUP signal (better a process at a time), this will generate some formatted information at a log file called info0001.log.
Of course, every node can be launched into a different computer, provided they are all into the same network segment (broadcast reaches all those computers). That's why they are separate instances...


Using the KinCLI (Kin Command Line Interface)

With the main kinode program, there's also the kincli application. It is still a testing application, but the intended use is to manage many of the cluster facilities and functions, getting status, etc. from an external command line interface. Just execute it, with one parameter (at least) giving the CM node IP (or host name). Type help and follow your instincts (help is a good choice if you don't know what to write at a given moment).

Return to home page from