Ok I removed the all guilty attributes from the config file and it does
not complain any more :
Now I get
myDB(admin) > show controllers
08:28:08,260 INFO controller.virtualdatabase.myDB
Member(address=ip-10-226-70-143.eu-west-1.compute.internal/10.226.70.143:7800,
uid=myDB) see members:[Member(address=/10.226.71.175:7800, uid=myDB),
Member(address=ip-10-226-70-143.eu-west-1.compute.internal/10.226.70.143:7800,
uid=myDB)] and has
mapping:{Member(address=ip-10-226-71-175.eu-west-1.compute.internal/10.226.71.175:7800,
uid=myDB)=10.226.71.175:1090,
Member(address=ip-10-226-70-143.eu-west-1.compute.internal/10.226.70.143:7800,
uid=myDB)=10.226.70.143:1090}
myDB is hosted by 2 controller(s):
10.226.71.175:1090
10.226.70.143:1090
Thank you all !
next step !
--
Damien Hardy
>
> Hi Robert,
>
> Thank for your help,
> I used the jgroups jar included in the sequoia distribution and start it
> as shown at
> http://www.jgroups.org/manual/html/user-advanced.html#IpNoMulticast
>
> My gossip seams to be running fine (I inspired me from the controler.sh
> script to create gossiprouter.sh) :
> ip-10-226-70-143:/usr/local/sequoia# bin/gossiprouter.sh
> GossipRouter is starting...
> log4j:WARN No appenders could be found for logger
> (org.jgroups.conf.ClassConfigurator).
> log4j:WARN Please initialize the log4j system properly.
> GossipRouter started at Thu Apr 16 07:25:19 UTC 2009
> Listening on port 12001 bound on address 0.0.0.0/0.0.0.0
>
> But when I start sequoia with my virtualdatabase (myDB) "<distribued/>",
> using the conf file you sent, I got (extract from a lot of other java
> errors):
>
> 07:34:20,711 ERROR jgroups.protocols.TCP the following properties are not
> recognized: {oob_thread_pool.enabled=false, thread_pool.enabled=false,
> thread_pool.rejection_policy=Run, thread_pool.queue_max_size=100,
> oob_thread_pool.min_threads=8, oob_thread_pool.keep_alive_time=5000,
> oob_thread_pool.rejection_policy=Run, thread_naming_pattern=cl,
> thread_pool.queue_enabled=false, thread_pool.max_threads=40,
> thread_pool.keep_alive_time=5000, oob_thread_pool.queue_max_size=100,
> use_concurrent_stack=false, oob_thread_pool.max_threads=20,
> thread_pool.min_threads=8, oob_thread_pool.queue_enabled=false}
> 07:34:20,714 ERROR controller.xml.DatabasesParser The virtual database
> myDB could not be added to the controller (because of Error while joining
> group myDB (org.continuent.hedera.channel.ChannelException: Exception when
> creating JGroups channel))
>
> I use sequoia-2.10.10-bin maybe it's a too old version and I have to use
> the dev one ?
>
> Cheers,
>
> --
> Damien
>
>
>> Hi Damien,
>>
>> To get things to work on EC2 you will probably need to set up a gossip
>> server using JGroups. Here's the general approach:
>>
>> 1.) Start an EC2 instance that will host the gossip server. Install
>> JGroups and start the gossip server as described in the JGroups
>> documentation.
>>
>> 2.) Configure Sequoia JGroups communication to point each Sequoia
>> controller to the gossip server. I have appended a typical
>> configuration
>> for this. You'll need to change HOST1, HOST2 to your gossip server host
>> names. For safety it's good to have a couple of them.
>>
>> 3.) Start the controllers and confirm they can connect to the group.
>>
>> The tricky thing on EC2 is to ensure that your controllers know the IP
>> address of the gossip server. One approach to simplify things is to
>> register that address through a dynamic DNS provider like
>> www.dnsmadeeasy.com. I have used this and it works pretty well.
>>
>> Cheers, Robert
>>
>>
>> On 4/15/09 8:41 AM PDT, "Damien Hardy" <[email protected]> wrote:
>>
>> Hi Emmanuel,
>>
>> many thanks for your help.
>> But have you some pointers on this kind of configuration ? (the UDP one
>> is
>> already quite obscure for me :)
>>
>> best regards
>>
>> --
>> Damien
>>
>>
>>>
>>> Hi Damien,
>>>
>>> AFAIK, Amazon EC2 does not support UDP multicast. You will have to use
>>> a
>>> TCP configuration of JGroups for the GMS (group membership) to work.
>>>
>>> Hope this helps,
>>> Emmanuel
>>>> I am playing with Sequoia for a couple of days now and I reach to have
>>>> a
>>>> complete chain working from php to 2 mysql databases working via
>>>> myosotis
>>>> and sequoia.
>>>>
>>>> Now 2nd round ... use 2 controllers.
>>>>
>>>> So I build a second VM with the same configuration of sequoia that my
>>>> first working one.
>>>>
>>>> it work again perfectly alone.
>>>>
>>>> No I try to make theme communicate as peer controllers.
>>>>
>>>> I add the
>>>> "<Distribution>
>>>> <MessageTimeouts/>
>>>> </Distribution>"
>>>> section in my virtualdatabase configuration file.
>>>> I rename the config/sequencer.xml and modify the
>>>> hedera_jgroups.properties
>>>> in consequences (in order to avoid the use of the sequencer.xml
>>>> present
>>>> in
>>>> the jgroups jar file).
>>>>
>>>> the controllers start without problem. But ...
>>>>
>>>> The can't see them each other so on the 2 controllers I obtain :
>>>> myDB(admin) > show controllers
>>>> 14:34:58,423 INFO controller.virtualdatabase.myDB
>>>> Member(address=/10.x.x.x:32781, uid=myDB) see
>>>> members:[Member(address=/10.x.x.x:32781, uid=myDB)] and has
>>>> mapping:{Member(address=/10.x.x.x:32781, uid=myDB)=10.x.x.x:1090}
>>>> myDB is hosted by 1 controller(s):
>>>> 10.226.71.175:1090
>>>>
>>>>
>>>> How does they discover themselves ? can I precise it ?
>>>>
>>>> I precise that I use 3 Amazon EC2 instances (virtual machines) for my
>>>> tests : 2 for db/controllers and 1 front (php)
>>>> So there are not on the same LAN I suppose.
>>>>
>>>> Thanks for your help in advance
>>>>
>>>> Best regards,
>>>>
>
>
> _______________________________________________
> Sequoia mailing list
> [email protected]
> http://forge.continuent.org/mailman/listinfo/sequoia
>
>
_______________________________________________
Sequoia mailing list
[email protected]
http://forge.continuent.org/mailman/listinfo/sequoia
