List Info

Thread: Default JGroups configs
Default JGroups configs
user name
 2007-10-24 12:04:41 
Any idea why the default configs still are:

  public String getDefaultClusterConfig()
    {
       return
"UDP(mcast_addr=224.0.0.36;mcast_port=55566;ip_ttl=32;&
quot; +
                
"mcast_send_buf_size=150000;mcast_recv_buf_size=80000):
" +
              
"PING(timeout=1000;num_initial_members=2):" +
              
"MERGE2(min_interval=5000;max_interval=10000):" +
               "FD_SOCK:" +
               "VERIFY_SUSPECT(timeout=1500):" +
               "pbcast.NAKACK 
(gc_lag=50;retransmit_timeout=600,1200,2400,4800):" +
              
"UNICAST(timeout=600,1200,2400,4800):" +
              
"pbcast.STABLE(desired_avg_gossip=20000):" +
               "FRAG(frag_size=8192):" +
              
"pbcast.GMS(join_timeout=5000;join_retry_timeout=2000;&
quot; +
              
"shun=false;print_local_addr=true):" +
               "pbcast.STATE_TRANSFER";
    }

instead of using STREAMING_STATE_TRANSFER and FLUSH instead
of  
STATE_TRANSFER?  Any good reasons not to change it?

Cheers,
--
Manik Surtani

Lead, JBoss Cache
JBoss, a division of Red Hat



_______________________________________________
jbosscache-dev mailing list
jbosscache-devlists.jboss.org
https://lists.jboss.org/mailman/listinfo/jbosscache-dev
Re: Default JGroups configs
user name
 2007-10-24 12:46:51 
Manik Surtani wrote:
> Any idea why the default configs still are:
> 
>  public String getDefaultClusterConfig()
>    {
>       return
"UDP(mcast_addr=224.0.0.36;mcast_port=55566;ip_ttl=32;&
quot; +
>              
"mcast_send_buf_size=150000;mcast_recv_buf_size=80000):
" +
>              
"PING(timeout=1000;num_initial_members=2):" +
>              
"MERGE2(min_interval=5000;max_interval=10000):" +
>               "FD_SOCK:" +
>               "VERIFY_SUSPECT(timeout=1500):"
+
>               
>
"pbcast.NAKACK(gc_lag=50;retransmit_timeout=600,1200,24
00,4800):" +
>              
"UNICAST(timeout=600,1200,2400,4800):" +
>              
"pbcast.STABLE(desired_avg_gossip=20000):" +
>               "FRAG(frag_size=8192):" +
>              
"pbcast.GMS(join_timeout=5000;join_retry_timeout=2000;&
quot; +
>              
"shun=false;print_local_addr=true):" +
>               "pbcast.STATE_TRANSFER";
>    }
> 
> instead of using STREAMING_STATE_TRANSFER and FLUSH
instead of 
> STATE_TRANSFER?  Any good reasons not to change it?
> 

Lots of other stuff missing vs. current best practice as
well. No FD. No FC.

Although, IMHO in 3.0 we should get rid of this altogether
and force 
people to provide a config.

-- 
Brian Stansberry
Lead, AS Clustering
JBoss, a division of Red Hat
brian.stansberryredhat.com
_______________________________________________
jbosscache-dev mailing list
jbosscache-devlists.jboss.org
https://lists.jboss.org/mailman/listinfo/jbosscache-dev
Re: Default JGroups configs
user name
 2007-10-24 12:52:50 
For 3.0, I agree.  Can't ditch it for 2.x though.

This came up analysing an intermittent failure in a state
transfer  
test and I found that FLUSH wasn't present!


On 24 Oct 2007, at 18:46, Brian Stansberry wrote:

> Manik Surtani wrote:
>> Any idea why the default configs still are:
>>  public String getDefaultClusterConfig()
>>    {
>>       return "UDP 
>>
(mcast_addr=224.0.0.36;mcast_port=55566;ip_ttl=32;" +
>>                
>>
"mcast_send_buf_size=150000;mcast_recv_buf_size=80000):
" +
>>              
"PING(timeout=1000;num_initial_members=2):" +
>>              
"MERGE2(min_interval=5000;max_interval=10000):" +
>>               "FD_SOCK:" +
>>              
"VERIFY_SUSPECT(timeout=1500):" +
>>               "pbcast.NAKACK 
>>
(gc_lag=50;retransmit_timeout=600,1200,2400,4800):" +
>>              
"UNICAST(timeout=600,1200,2400,4800):" +
>>              
"pbcast.STABLE(desired_avg_gossip=20000):" +
>>               "FRAG(frag_size=8192):" +
>>               "pbcast.GMS 
>> (join_timeout=5000;join_retry_timeout=2000;"
+
>>              
"shun=false;print_local_addr=true):" +
>>               "pbcast.STATE_TRANSFER";
>>    }
>> instead of using STREAMING_STATE_TRANSFER and FLUSH
instead of  
>> STATE_TRANSFER?  Any good reasons not to change
it?
>
> Lots of other stuff missing vs. current best practice
as well. No  
> FD. No FC.
>
> Although, IMHO in 3.0 we should get rid of this
altogether and  
> force people to provide a config.
>
> -- 
> Brian Stansberry
> Lead, AS Clustering
> JBoss, a division of Red Hat
> brian.stansberryredhat.com

--
Manik Surtani

Lead, JBoss Cache
JBoss, a division of Red Hat



_______________________________________________
jbosscache-dev mailing list
jbosscache-devlists.jboss.org
https://lists.jboss.org/mailman/listinfo/jbosscache-dev
Re: Default JGroups configs
user name
 2007-10-24 19:06:17 
Also, should the default be TCP with MPING rather than UDP?

On 24 Oct 2007, at 18:52, Manik Surtani wrote:

> For 3.0, I agree.  Can't ditch it for 2.x though.
>
> This came up analysing an intermittent failure in a
state transfer  
> test and I found that FLUSH wasn't present!
>
>
> On 24 Oct 2007, at 18:46, Brian Stansberry wrote:
>
>> Manik Surtani wrote:
>>> Any idea why the default configs still are:
>>>  public String getDefaultClusterConfig()
>>>    {
>>>       return "UDP 
>>>
(mcast_addr=224.0.0.36;mcast_port=55566;ip_ttl=32;" +
>>>                
>>>
"mcast_send_buf_size=150000;mcast_recv_buf_size=80000):
" +
>>>              
"PING(timeout=1000;num_initial_members=2):" +
>>>              
"MERGE2(min_interval=5000;max_interval=10000):" +
>>>               "FD_SOCK:" +
>>>              
"VERIFY_SUSPECT(timeout=1500):" +
>>>               "pbcast.NAKACK 
>>>
(gc_lag=50;retransmit_timeout=600,1200,2400,4800):" +
>>>              
"UNICAST(timeout=600,1200,2400,4800):" +
>>>              
"pbcast.STABLE(desired_avg_gossip=20000):" +
>>>               "FRAG(frag_size=8192):"
+
>>>               "pbcast.GMS 
>>>
(join_timeout=5000;join_retry_timeout=2000;" +
>>>              
"shun=false;print_local_addr=true):" +
>>>              
"pbcast.STATE_TRANSFER";
>>>    }
>>> instead of using STREAMING_STATE_TRANSFER and
FLUSH instead of  
>>> STATE_TRANSFER?  Any good reasons not to change
it?
>>
>> Lots of other stuff missing vs. current best
practice as well. No  
>> FD. No FC.
>>
>> Although, IMHO in 3.0 we should get rid of this
altogether and  
>> force people to provide a config.
>>
>> -- 
>> Brian Stansberry
>> Lead, AS Clustering
>> JBoss, a division of Red Hat
>> brian.stansberryredhat.com
>
> --
> Manik Surtani
>
> Lead, JBoss Cache
> JBoss, a division of Red Hat
>
>
>
> _______________________________________________
> jbosscache-dev mailing list
> jbosscache-devlists.jboss.org
> https://lists.jboss.org/mailman/listinfo/jbosscache-dev

--
Manik Surtani

Lead, JBoss Cache
JBoss, a division of Red Hat



_______________________________________________
jbosscache-dev mailing list
jbosscache-devlists.jboss.org
https://lists.jboss.org/mailman/listinfo/jbosscache-dev
Re: Default JGroups configs
user name
 2007-10-25 02:34:58 
   1.   I'd actually return "udp.xml", or at least
a ref to some XML
      file that ships with JGroups, rather than hard-code
that into a string
   2. If you use STREAMING_STATE_TRANSFER, then this
requires the user
      to implement the right callbacks, which I don't think
many folks
      do. IIRC, Vladimir uses the byte[] buffer methods if
that's not
      done, but I might be wrong


Manik Surtani wrote:
> Any idea why the default configs still are:
>
>  public String getDefaultClusterConfig()
>    {
>       return
"UDP(mcast_addr=224.0.0.36;mcast_port=55566;ip_ttl=32;&
quot; +
>              
"mcast_send_buf_size=150000;mcast_recv_buf_size=80000):
" +
>              
"PING(timeout=1000;num_initial_members=2):" +
>              
"MERGE2(min_interval=5000;max_interval=10000):" +
>               "FD_SOCK:" +
>               "VERIFY_SUSPECT(timeout=1500):"
+
>               
>
"pbcast.NAKACK(gc_lag=50;retransmit_timeout=600,1200,24
00,4800):" +
>              
"UNICAST(timeout=600,1200,2400,4800):" +
>              
"pbcast.STABLE(desired_avg_gossip=20000):" +
>               "FRAG(frag_size=8192):" +
>              
"pbcast.GMS(join_timeout=5000;join_retry_timeout=2000;&
quot; +
>              
"shun=false;print_local_addr=true):" +
>               "pbcast.STATE_TRANSFER";
>    }
>
> instead of using STREAMING_STATE_TRANSFER and FLUSH
instead of 
> STATE_TRANSFER?  Any good reasons not to change it?
>
>

-- 
Bela Ban
Lead JGroups / Clustering Team
JBoss - a division of Red Hat
_______________________________________________
jbosscache-dev mailing list
jbosscache-devlists.jboss.org
https://lists.jboss.org/mailman/listinfo/jbosscache-dev
Re: Default JGroups configs
user name
 2007-10-25 07:14:58 
Bela Ban wrote:
>   1.   I'd actually return "udp.xml", or at
least a ref to some XML
>      file that ships with JGroups, rather than
hard-code that into a 
> string
Yes it is definitely better to reference jgroups conf file.
Either 
udp.xml or flush-udp.xml is ok.

>   2. If you use STREAMING_STATE_TRANSFER, then this
requires the user
>      to implement the right callbacks, which I don't
think many folks
>      do. IIRC, Vladimir uses the byte[] buffer methods
if that's not
>      done, but I might be wrong

No need for special callbacks. This is completely
transparent. User can 
switch back and forth between these two types of transfer
without any 
additional coding. We'll do that in JGroups 3.0 as well.





_______________________________________________
jbosscache-dev mailing list
jbosscache-devlists.jboss.org
https://lists.jboss.org/mailman/listinfo/jbosscache-dev
[1-6]

about | contact  Other archives ( Real Estate discussion Medical topics )