Re: [javagroups-users] Problems with multicast on many machines

[Jan B. <js...@gm...>]

Thanks Bela!

That helped a lot, we are upgrading from 2.2.7 so it was time to look at 
the stack too ;)

Jan S

Bela Ban wrote:
> I see you use CAUSAL, which we don't support (as a matter of fact, it 
> was removed in 2.5). Also, you're misisng max_bytes in STABLE.  On top 
> of that, there is no flow control in your stack: flow control is 
> essential if you're sending lots of messages !
> I recommend use udp.xml which ships with 2.4 and modify it slightly, 
> e.g. num_initial_members in PING etc.
>
> Your description is quite vague, but the above tips might solve your 
> problem.
>
> Jan Berg wrote:
>> Hi all,
>>
>> I have some problems using JGroups multicast on a 16 machine setup.
>>
>> During a stress test of the system most messages comes through, but
>> sometimes it seems like multicast messages does not get through for some
>> time. When I fallback to sending unicast to the members not receiving
>> the multicast messages they get them immediately.
>>
>> After two minutes the multicast messages comes in a burst
>> (Is this a timeout somewhere in JGroups? or is it my router?)
>> or sometimes I even seem to loose the messages
>>
>> I have tried using using different versions of JGroups (2.4.0, 2.4.1sp4
>> and 2.5) and also different JDKs (1.5 and 1.6), and I get the best
>> results using 2.4.0 and JDK 1.6...
>>
>> Here's my stack for 2.4:
>> UDP(mcast_addr=" + mcast_addr + ";" +
>> "mcast_port=" + mcast_port + ";" +
>> "ip_ttl=32;bind_addr=" + bind_addr + ";" +
>> "mcast_send_buf_size=450000;" +
>> "mcast_recv_buf_size=3600000):" +
>> "PING(timeout=2000;num_initial_members=10):" +
>> "MERGE2(min_interval=5000;max_interval=10000):" +
>> "FD:" +
>> "VERIFY_SUSPECT(timeout=1500):" +
>> "pbcast.NAKACK(max_xmit_size=8192;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;down_thread=false;up_thread=false):CAUSAL:" +
>> "pbcast.GMS(join_timeout=5000;join_retry_timeout=2000;" +
>> "shun=false;print_local_addr=false)" +
>> ":pbcast.STATE_TRANSFER";
>>
>> Any ideas for what could be wrong here?
>