GPnP defines a node’s metadata, including information about network interfaces for public and private
connections, the ASM server parameter file, and CSS voting disks.

Cluster Startup Sequence

• Start Database Machine
• Start OHASD Service
• Start GPnP Daemon
• Start Oracle ASM

Oracle Cluster Startup Sequence first start Node or Machine then start OHASD Service , third Start GPNP Daemon and fourth start Oracle ASM.

As We Know OCR and Voting Disk are on ASM, but Clusterware requires them to start CRSD and CSSD, and both of these files are on ASM, which itself is a resource for the node. So, how does Clusterware start? Which files does it access to get the information needed to initiate Clusterware?
To address this, Oracle introduced two local operating system files: OLR and GPnP.

When a node in an Oracle Clusterware cluster restarts, OHASD starts using methods specific to the platform at that stage OHASD will accesses the OLR (Oracle Local Registry) which stored on the local file system. The OLR provides the necessary information to complete the OHASD initialization.

OHASD starts the GPnP Daemon and CSS Daemon. The CSS Daemon will access
the GPnP Profile which stored on the local file system.

The GPnP service is a distributed system where all GPnP agents work together to replicate and share configuration profiles across a network.

Each node in the network runs its own GPnP agent, and these agents communicate with each other in a peer-to-peer manner, meaning there’s no central or master agent which controlling the process.

This design ensures high availability because even if one GPnP agent fails, the other agents continue to function, allowing new nodes to join and access the necessary profiles.

GPnP uses the standard IP multicast protocol, provided by mDNS (Multicast DNS), to discover and locate other GPnP agents within the network. Through this multicast discovery, a GPnP agent on a new node can automatically find other agents that may have the profiles it needs, without requiring manual configuration. This makes the GPnP service resilient and self-managing in a clustered environment.

Example

The GPnP service is like a group of friends who all share important information with each other. Each friend (or GPnP agent) has a copy of this information, and they work together to keep everything up to date.

If one friend can’t help because they’re busy (or their computer crashes), the other friends can still share the information with new people who join the group. This makes sure that the group can keep working even if one person isn’t available.

The way these friends find each other is like using a special kind of message that everyone can hear (this is called multicast). When a new friend joins the group, they use this special message to find the others and get the information they need, without having to ask for directions or set anything up ahead of time.

So, in short, GPnP is a system that helps computers in a network share important information with each other automatically, and it keeps working smoothly even if one computer has a problem.