Oh I see, I was thinking node specific state. Yes things like IP address, working address etc can be marked with some prefix like _local and then weed out using filters.
Rajith On 5/18/07, Afkham Azeez <[EMAIL PROTECTED]> wrote:
Generally, in the clustered scenario, all nodes will share the same URL repo. Nodes may store node specifi information e.g. working directory, local IP address etc. Nobody can impose restrictions saying that node specific data cannot be maintained. Azeez On 5/18/07, Rajith Attapattu <[EMAIL PROTECTED]> wrote: > > Dims, Afkam, > > Dims suggestion is good. I agree that there can be static data that > comes from configuration and doesn't need to be replicated. > So weedng them out is useful. > However having node specific data is still not a good idea IMO (even > though u can weed them out from replication using Dims idea). > One example given by Chamikara is node specific policies from sandesha > policy bean. I think u need identical policies (and other data) for replicas > in a cluster. > Afkam can u provide some examples for node specific proeprties ? (Maybe > I didn't understand your point) > > Regards, > > Rajith > > On 5/18/07, Davanum Srinivas < [EMAIL PROTECTED]> wrote: > > > > Only other thing i can think of is similar to you know the proxy > > settings...have regexp based includes and excludes (on the key) in the > > ClusterManager for the default properties we already have and allow > > API access (and/or axis2.xml entries) for folks to add/delete from the > > list of includes/excludes. that's that other extreme... > > > > -- dims > > > > On 5/18/07, Davanum Srinivas < [EMAIL PROTECTED] > wrote: > > > Azeez, > > > > > > Did you already rule out a simple solution? If a service author > > wants > > > a specific custom property to be available, then they can add a > > simple > > > prefix to the key in the Map? > > > > > > thanks, > > > dims > > > > > > On 5/18/07, Afkham Azeez <[EMAIL PROTECTED]> wrote: > > > > We have a problem when it comes to replicating properties in our > > clustering > > > > implementation. There are some properties which are specific to a > > node, > > > > specially properties in the ConfigurationContext. Some properties > > are added > > > > by different modules such as Sandesha2, Rampart to the > > ConfigurationContext. > > > > One thing is that these objects are not serializable, and the > > other thing is > > > > that these properties should not be replicated. Some information > > which are > > > > specific to a node may be added to the ConfigurationContexts, and > > these > > > > should never be replicated. > > > > > > > > So there should be some way to inform Axis2 about the properties > > that need > > > > to be clustered and those that shouldn't be clustered. > > > > > > > > I suggest we introduce a new Map to AbstractContext called > > > > clusterableProperties. Stuff that are added to this Map will be > > replicated, > > > > and the service author/module author has to add the properties > > that need to > > > > be replicated into the clusterableProperties Map. > > > > > > > > Thoughts? > > > > > > > > -- > > > > Thanks > > > > Afkham Azeez > > > > > > > > http://www.wso2.org > > > > GPG Fingerprint: 643F C2AF EB78 F886 40C9 B2A2 4AE2 C887 665E > > 0760 > > > > > > > > > -- > > > Davanum Srinivas :: http://davanum.wordpress.com > > > > > > > > > -- > > Davanum Srinivas :: http://davanum.wordpress.com > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > > -- Thanks Afkham Azeez http://www.wso2.org GPG Fingerprint: 643F C2AF EB78 F886 40C9 B2A2 4AE2 C887 665E 0760
