stuck in dependency cycle on clusters/datasource

[mdplatts]

Probably me doing it wrong and I apologise if this is the wrong place, but I'm getting the following error -

Error: Failed to apply catalog: Found 1 dependency cycle: (Class[Datasources] => Wls_datasource[xxx/ClusterLeasingDataSource] => Wls_cluster[xxx/NameServerCluster] => Class[Cluster] => Class[Datasources])

I have this data source -

datasource_instances:
  xxx/ClusterLeasingDataSource:
    ensure: present
    jndinames: ClusterLeasingDataSource
    drivername: oracle.jdbc.driver.OracleDriver
    globaltransactionsprotocol: None
    usexa: '0'
    initialcapacity: '2'
    maxcapacity: 50
    testtablename: 'SQL SELECT 1 FROM DUAL'
    target:
      - NameServerCluster
    targettype:
      - Cluster

And this cluster definition -

cluster_instances:
  xxx/NameServerCluster:
    ensure: present
    messagingmode:              'unicast'
    #migrationbasis:             'database'
    migrationbasis:             'consensus'
    #migration_datasource:       'ClusterLeasingDataSource'
    servers:
      - 'ns_1'
      - 'ns_2'
      - 'ns_3'

If I leave it as 'consensus' then puppet runs but if I change it to "database" and "ClusterLeasingSource" then I get the dependency error. I'm trying to do that to see if I can get a 3-node cluster to run as a one node cluster (e.g. in the event of the failure of two of the three AZ's in our AWS account). I can reduce 3 to 2 nodes but not 2 to 1 (or even reboot the one) and someone suggested I try switching to database.

I'd also want to use the data source for mutliple different clusters - which given the "migration_table_name" value I'm presuming it should be possible to share a schema and just have several tables in the one place.

[biemond]

Hi

you got the chicken and egg situations but you can always assign the datasource to the cluster nodes (servers) instead of cluster. That also works.

thanks

[mdplatts]

Thanks - I tried that and it still reports the same as if its ignored the changes (unless I misunderstood which is always an option) -

datasource_instances:
  xxx/ClusterLeasingDataSource:
    ensure: present
    jndinames: ClusterPool
    drivername: oracle.jdbc.driver.OracleDriver
    globaltransactionsprotocol: None
    usexa: '0'
    initialcapacity: '2'
    maxcapacity: '50'
    testtablename: 'SQL SELECT 1 FROM DUAL'
    target:
      - ns_1
      - ns_2
      - ns_3
    targettype:
      - Server
      - Server
      - Server

The graph remains the same as if its ignored the changes -

digraph Resource_Cycles {
  label = "Resource Cycles"
"Class[Datasources]" -> "Wls_datasource[strata/ClusterLeasingDataSource]" -> "Wls_cluster[xxx/EchCluster]" -> "Class[Cluster]" -> "Class[Datasources]"
"Class[Datasources]" -> "Wls_datasource[strata/ClusterLeasingDataSource]" -> "Wls_cluster[xxx/NameServerCluster]" -> "Class[Cluster]" -> "Class[Datasources]"
"Class[Datasources]" -> "Wls_datasource[strata/ClusterLeasingDataSource]" -> "Wls_cluster[xxx/SICluster]" -> "Class[Cluster]" -> "Class[Datasources]"
"Class[Datasources]" -> "Wls_datasource[strata/ClusterLeasingDataSource]" -> "Wls_cluster[xxx/XxxCluster]" -> "Class[Cluster]" -> "Class[Datasources]"
"Class[Datasources]" -> "Wls_datasource[strata/ClusterLeasingDataSource]" -> "Wls_cluster[xxx/SupCluster]" -> "Class[Cluster]" -> "Class[Datasources]"
}

[biemond]

Ok,

but now I think it is in your code , you added some dependency on class Cluster back to datasources again.

[mdplatts]

Its just as in the first post - but with the database options uncommented to "enable" the feature. e.g.

cluster_instances:
  xxx/NameServerCluster:
    ensure: present
    messagingmode:               'unicast'
    migrationbasis:                  'database'
    migration_datasource:       'ClusterLeasingDataSource'
    migration_table_name:      'XXX_NS' 
    servers:
      - 'ns_1'
      - 'ns_2'
      - 'ns_3'

Will take another look - thanks.

[biemond]

if I look at this , it should be first servers -> datasource -> cluster.

can you show me your manifest and the hiera parts.

[mdplatts]

Sorry for the delay - not had chance to try this yet (I did a quick test but it didnt seem to fix it but it might still be me).

A quick question rather than a separate log (unless you or github would prefer it to be separate) - is there support for startup classes. I cannot see it but I need to add some to support a 3rd party instrumentation/monitoring tool - CA APM (https://docops.ca.com/ca-apm/10-1/en/implementing-agents/java-agent/install-the-java-agent/configure-application-server-to-use-the-java-agent/oracle-weblogic/)

I can add startup arguments and extra jar files to the classpath but I've not seen that a startupclass is supported.

I support I might need to run a wlst script to do it or try (unlikely to succeed) to add it to the module.

[biemond]

Hi,

I guess I don't support this but you can send me a PullRequest

Thanks

[mdplatts]

will see what I can do.