Installing Scalaris Problem

[Nika Abramova]

Greetings,

I am currently benchmarking some NoSQL databases using YCSB. So I found Scalaris layer and decided to try it.

I was following guide to install scalaris and I downloaded it from https://scalaris.googlecode.com/files/scalaris-0.5.0.tar.gz and then ./configure, make & make docs. However, I don't have boot.sh and other scripts it said in guide. 

Files I have at /bin are: find_erlang.bat, firstnode.bat, forstnode.sh, joining_node.bat,joining_node.sh, mgmt-server.sh, quorum_node.sh, scalaris.cfg, scalarisctl.conf, scalariscrl.in and scalaris.local.cfg.example.

Then I tried to tar again and after getting into bin folder are all same files.

How should I execute?

Regards, Veronika

[Nico Kruber]

Hi Veronika,
I am wondering which guide you are following - I scanned quickly through our
documentation and nothing points to the old boot.sh script any more (except
one code excerpt in the user-dev-guide).

Anyway, firstnode.* is the replacement for the old boot.sh (.bat for Windows,
.sh for Linux/Unix etc.)

This may help you for a quick-start:
https://code.google.com/p/scalaris/wiki/FAQ

The full guide should be in the user-dev-guide folder of your checkout or
extracted .tar file.

And I guess you probably found this one:
https://code.google.com/p/scalaris/wiki/YCSB

Anyway, please keep in mind that Scalaris offers full ACID transactions and as
such may not be comparable with other data stores by the raw numbers that YCSB
exports. Transactions come with a cost!


Regards
Nico

[Nika Abramova]

Hello,

you are right, I was following "Users and Developers Guide" at http://eucalyptus.linux4u.jp/wiki/index.php?plugin=attach&refer=Scalaris%2F%E3%82%A4%E3%83%B3%E3%82%B9%E3%83%88%E3%83%BC%E3%83%AB%E6%96%B9%E6%B3%95&openfile=main.pdf since I searched for quick install and that was first result and I was expecting it to work.

Now I have started with firstnode.sh but when I try to run YCSB over it I am getting: 

de.zib.scalaris.ConnectionException: Cannot connect to peer node, bad nodes: [node1@ubuntu], good nodes: [], retries: 3

I can not run just 1 instance? I have to always run it as cluster by executing ./bin/joining_node.sh, ./bin/joining_node.sh 2 and ./bin/joining_node.sh 3 ?

[Nico Kruber]

On Wednesday 24 Jul 2013 02:32:03 Nika Abramova wrote:
> Hello,
> you are right, I was following "Users and Developers Guide" at
> http://eucalyptus.linux4u.jp/wiki/index.php?plugin=attach&refer=Scalaris%2F%
> E3%82%A4%E3%83%B3%E3%82%B9%E3%83%88%E3%83%BC%E3%83%AB%E6%96%B9%E6%B3%95&open

> file=main.pdf since I searched for quick install and that was first result

> and I was expecting it to work.

our project site at https://code.google.com/p/scalaris/ has a correct link to
the "Users and Developers Guide" (and as I said, it's in the tarball as well)

> Now I have started with firstnode.sh but when I try to run YCSB over it I
> am getting:
> de.zib.scalaris.ConnectionException: Cannot connect to peer node, bad
> nodes: [node1@ubuntu], good nodes: [], retries: 3

if you start via firstnode.sh then your node will probably be named
"firstnode@ubuntu" and you need to adapt the YCSB connection settings
accordingly:
scalaris.node = "firstnode@ubuntu"
(you could also use firstnode@localhost in which case "localhost" will be
replaced by the correct hostname)

see
https://code.google.com/p/scalaris/wiki/YCSB

> I can not run just 1 instance? I have to always run it as cluster by
> executing ./bin/joining_node.sh, ./bin/joining_node.sh 2 and ./bin/
> joining_node.sh 3 ?

you can run a single instance - the 4 replicas that are stored for each item
you add are used nonetheless (on this single node)!
-> I'd recommend this for debugging only as it is a rather un-common setup for
a distributed data store. A normal setup would of course be more expensive as
communication between the nodes is needed.
Also a difference you have to keep in mind when comparing different data
stores! (distributed vs. single-node data stores)


Nico

[Nika Abramova]

Got it! I am going to research and study more about Hypertable in order to understand those differences better.

Thank you very much for help, that was my mistake for not reading troubleshooting part. That was because I thought that Hypertable must be similar to Riak, where in order to test it it is necessary to run a cluster.

I hope I won't get any future errors.

Once again, thank you for your help and time.

Regards, Veronika