Hello Alexander,

Under the covers, 71ML7 incorporates two big architectural changes - elastic search and hazelcast.

Traditionally, PRPC clustering was 'loosely coupled' - PRPC nodes did not communicate directly with each other by default; they simply shared the database. The only cross-node communication normally implemented were SOAP calls to the Lucene Index owning node, and the SOAP calls were often set using a proxy or load balance URL. PRPC did not open ports or directly talk cross-node.

In ML7, PRPC clusters are tightly coupled.

a- Lucene search has been replaced by elastic search. With elastic search, the indexing nodes open a communication port for communicate with both index-client and other indexer nodes.

b- Hazelcast has been added as a core engine component. It is currently used for DSM and agent management but later will replace use of database for lock management. Each PRPC node opens a unique port for hazelcast communication to other PRPC nodes.

By default, elastic looks for an available listen port in 9300-9399 range. You must validate that ports 9300-9399 are unblocked / available. There are system settings to override the elastic port range.

I believe Hazelcast looks for an available port in the 5700-5800 range (have not verified). You must validate that ports 5700-5800 are unblocked / available. I believe there are system settings to override the hazelcast port range.

PRPC servers must be able to cross-communicate via TCP/IP. There will likely be firewall changes needed in production and cloud environments. I do not believe that PegaCloud has the hazelcast and elastic port blocks open by default.

Two new columns are added to pr_sys_Statusnodes - pyclusteraddress holds the hazelcast IP address/port, pyindexeraddress holds the elastic address/port.

Kindly let me know if this information might helped you.

Regards,

Mahesh

Hello Alexander,

Under the covers, 71ML7 incorporates two big architectural changes - elastic search and hazelcast.

Traditionally, PRPC clustering was 'loosely coupled' - PRPC nodes did not communicate directly with each other by default; they simply shared the database. The only cross-node communication normally implemented were SOAP calls to the Lucene Index owning node, and the SOAP calls were often set using a proxy or load balance URL. PRPC did not open ports or directly talk cross-node.

In ML7, PRPC clusters are tightly coupled.

a- Lucene search has been replaced by elastic search. With elastic search, the indexing nodes open a communication port for communicate with both index-client and other indexer nodes.

b- Hazelcast has been added as a core engine component. It is currently used for DSM and agent management but later will replace use of database for lock management. Each PRPC node opens a unique port for hazelcast communication to other PRPC nodes.

By default, elastic looks for an available listen port in 9300-9399 range. You must validate that ports 9300-9399 are unblocked / available. There are system settings to override the elastic port range.

I believe Hazelcast looks for an available port in the 5700-5800 range (have not verified). You must validate that ports 5700-5800 are unblocked / available. I believe there are system settings to override the hazelcast port range.

PRPC servers must be able to cross-communicate via TCP/IP. There will likely be firewall changes needed in production and cloud environments. I do not believe that PegaCloud has the hazelcast and elastic port blocks open by default.

Two new columns are added to pr_sys_Statusnodes - pyclusteraddress holds the hazelcast IP address/port, pyindexeraddress holds the elastic address/port.

Kindly let me know if this information might helped you.

Regards,

Mahesh