Hello, We came across a potential problem in the clustering module. According to the JGroups documentation:
"When a network error occurs, the cluster might be partitioned into several different partitions. JGroups has a MERGE service that allows the coordinators in partitions to communicate with each other and form a single cluster back again." In our environment this indeed resulted to a problem in the case of two coordinators sharing one channel. Costyn van Dongen van IC&S found a way to prevent it. With the change described here, two coordinators will be merged and 1 coordinator will be selected. It would be nice to have this change in the CVS. What is the procedure to have this change accepted? (I suppose it should be a VOTE CALL, but this should be then initiated by a MMC member, am I right?) Anyway, here is a description of the change: The JGroups protocol stack is defined in multicast.xml (in 1.7.1 version) en jgroup.xml (in the current cvs/1.8). In the current CVS, this is the file I am talking about: http://cvs.mmbase.org/viewcvs/*checkout*/applications/clustering/config/ utils/jgroup.xml?content-type=text%2Fplain By adding the following line between the lines of PING en FD, the problem should be prevented: MERGE2(min_interval=5000;max_interval=10000). The file then becomes: <property name="channelproperties"> UDP(mcast_addr=224.0.0.1;mcast_port=16080;ip_ttl=1): PING(timeout=3000;num_initial_members=6): MERGE2(min_interval=5000;max_interval=10000): FD(timeout=3000): VERIFY_SUSPECT(timeout=1500): pbcast.NAKACK(gc_lag=10;retransmit_timeout=600,1200,2400,4800): UNICAST(timeout=600,1200,2400,4800): pbcast.STABLE(desired_avg_gossip=10000): FRAG: pbcast.GMS(join_timeout=5000;join_retry_timeout=2000;shun=true;print_loc al_addr=true) </property> Greetings, Nadia Poulou Stichting Kennisnet DISCLAIMER: Dit bericht (met bijlagen) is met grote zorgvuldigheid samengesteld. Voor mogelijke onjuistheid en/of onvolledigheid van de hierin verstrekte informatie kan Kennisnet geen aansprakelijkheid aanvaarden, evenmin kunnen aan de inhoud van dit bericht (met bijlagen) rechten worden ontleend. De inhoud van dit bericht (met bijlagen) kan vertrouwelijke informatie bevatten en is uitsluitend bestemd voor de geadresseerde van dit bericht. Indien u niet de beoogde ontvanger van dit bericht bent, verzoekt Kennisnet u dit bericht te verwijderen, eventuele bijlagen niet te openen en wijst Kennisnet u op de onrechtmatigheid van het gebruiken, kopiëren of verspreiden van de inhoud van dit bericht (met bijlagen). This message (with attachments) is given in good faith. Kennisnet cannot assume any responsibility for the accuracy or reliability of the information contained in this message (with attachments), nor shall the information be construed as constituting any obligation on the part of Kennisnet. The information contained in this message (with attachments) may be confidential or privileged and is only intended for the use of the named addressee. If you are not the intended recipient, you are requested by Kennisnet to delete this message (with attachments) without opening it and you are notified by Kennisnet that any disclosure, copying or distribution of the information contained in this message (with attachments) is strictly prohibited and unlawful. _______________________________________________ Developers mailing list [email protected] http://lists.mmbase.org/mailman/listinfo/developers
