Thanks Karl, There is a deviation here, all MCF tomcats & processes are not pointing to same properites.xml, instead we have two properties.xml under /multiprocess-zk-example folders on each of MCF setup. So for node 1 all zk nodes are updated in connectstring parameter & for node 2, all zk nodes are updated in connectstring parameter.
All six zookeeper nodes know about each other as i have updated for all six nodes in zoo.cfg for each of zk nodes. I have configured two set of wars on two different tomcats on separate physical machines for cluster setup. Also as per your comment. If you are going to use two tomcats, each one must have a -D switch pointing to the same properties.xml as used by the agents processes Should i use only one properties.xml for both nodes/setup? Please suggest. Regards. On Mon, Sep 1, 2014 at 1:06 PM, Karl Wright <[email protected]> wrote: > Hi Lalit, > > So, just to clarify: > - All MCF tomcats and processes that you are describing share the same > properties.xml > - That properties.xml describes six zookeeper instances > > Questions: > > 1) Do those six zookeeper instances each know about the other five? They > should; it should be one group of six, rather than two groups of 3. > 2) It's not clear to me why you are trying to have two sets of wars > running, although this should work. The war files are not responsible for > managing data, just giving you access to it. Can you clarify? > > If you are going to use two tomcats, each one must have a -D switch > pointing to the same properties.xml as used by the agents processes. If > that is done, I can think of no reason why both would not work at the same > time -- although bear in mind that ManifoldCF does not have any facility to > prevent unhappy results if two people or more are editing the same thing at > the same time in the crawler UI or via the API. > > Thanks, > Karl > > > > > > On Mon, Sep 1, 2014 at 1:02 AM, lalit jangra <[email protected]> > wrote: > >> Thanks Karl, >> >> I am running all three wars: mcf-api-service.war, >> mcf-authority-service.war & mcf-crawler-ui.war on tomcat. In properties.xml >> under /multiprocess-zk-example for >> "org.apache.manifoldcf.zookeeper.connectstring" property, i have pointed to >> all six zookeepers in both nodes. >> >> <property name="org.apache.manifoldcf.zookeeper.connectstring" >> value="server1zk:2181,server1k:2182,server1zk:2183,server2zk:2181,server2zk:2182,server2zk:2183"/> >> >> So all zk processes point to same cluster. I have a single cluster where >> there are two tomcats at two different nodes & each having 3 zk processes. >> These both setups point to same six zk processes and in >> properties-global.xml of both MCF i have configured to point out to same >> PostgreSQL DB. >> >> >> On Sun, Aug 31, 2014 at 11:56 PM, Karl Wright <[email protected]> wrote: >> >>> bq. but i could observe that if i start all three zookeepers on say >>> first node and start MCF using tomcat along with agent on it, i am not able >>> to access MCF admin UI here. Ideally i should be able to access admin UI as >>> per Zookeper's 50% rule. Is there something i am missing? >>> >>> Can you be more specific? What war(s) are you running on tomcat? Do >>> all processes belonging to "node 1" point to the same properties.xml? >>> >>> If you are trying to set up a CLUSTER, all processes in the cluster >>> should point to the same properties.xml and thus share the same zookeeper >>> members. If you are trying to have TWO clusters, one for (say) backup, >>> then they have two DIFFERENT properties.xml files, two different sets of >>> Zookeeper machines, and two different database instances. >>> >>> >>> Karl >>> >>> >>> >>> On Sun, Aug 31, 2014 at 12:36 PM, lalit jangra <[email protected] >>> > wrote: >>> >>>> Hi, >>>> >>>> I have configured two MCF instances on two separate tomcats on >>>> different servers with each having three zookeeper instance (total six) >>>> having one common PostgreSQL DB. Its running fine but i could observe that >>>> if i start all three zookeepers on say first node and start MCF using >>>> tomcat along with agent on it, i am not able to access MCF admin UI here. >>>> Ideally i should be able to access admin UI as per Zookeper's 50% rule. Is >>>> there something i am missing? >>>> >>>> Secondly i want to confirm once i make whole setup up and running i.e. >>>> all six zookeepers running along with both MCFs on tomcat as well as agent >>>> 1 on node 1 and agent 2 on node 2 and i start to crawl a source say >>>> SharePoint, is that both agents crawl the source parallely (as these are >>>> pointing to same DB instance) and hence i could see double of processing >>>> here? Or how agents behave in this case? >>>> >>>> Please suggest. >>>> >>>> Regards, >>>> Lalit. >>>> >>> >>> >> >> >> -- >> Regards, >> Lalit. >> > > -- Regards, Lalit.
