How do I set up PI Data Archive high availability with collectives?
How do I set up PI Data Archive high availability with collectives?
A PI Collective is a group of replicated PI Data Archive servers that provide high availability and load distribution.
Architecture
A collective consists of:
- One primary member — receives all writes
- One or more secondaries — receive replicated data from the primary
- Clients connect to the collective name, not individual servers
How Replication Works
- Data arrives at the primary via interfaces and PI Buffer Subsystem
- The primary writes to its archive and snapshot
- PItoPI interface on each secondary pulls data from the primary
- Secondaries maintain their own copies of archive files
Setup Steps
1. Install PI Data Archive on All Nodes
Each member needs its own PI Data Archive installation with matching:
- PI point configuration (auto-synced)
- Archive settings
- Security configuration
2. Create the Collective
Use PI System Management Tools (SMT) > Collectives:
- Register the primary server
- Add secondary members
- Configure replication priority and network settings
3. Configure PItoPI
The PItoPI interface handles data replication:
- Runs on each secondary
- Connects to the primary's snapshot and archive
- Handles catch-up after outages
4. Set Up Buffering
PI Buffer Subsystem on interface nodes should be configured with the collective:
- Knows all collective members
- Automatically redirects writes during failover
Failover Behavior
| Scenario | Behavior |
|---|---|
| Primary goes down | Secondaries continue serving reads; writes buffer on interface nodes |
| Primary comes back | Buffered data replays; replication resumes |
| Secondary goes down | No impact on writes; reduced read capacity |
| Network split | Buffer Subsystem queues data until connectivity restores |
Best Practices
- Place primary and secondary in different physical locations
- Keep all collective members on the same PI version
- Monitor replication lag — a growing gap indicates a problem
- Test failover regularly — don't wait for a real outage to discover issues
Want to ask a follow-up?
PiChat can help with your specific PI System use case. Ask follow-up questions, get code examples, and troubleshoot issues.
Related Questions
Dive Deeper
More Questions
How do I configure PI AF templates?How do I authenticate with PI Web API from Python?What's the difference between snapshot and archive values in PI?What's the difference between PI Vision and PI ProcessBook?How do I build effective PI Vision displays?What are PI Event Frames and how do I use them?How do I write PI AF Analytics expressions?How do I use PI DataLink in Excel?Should I use PI Web API or AF SDK for my application?How do I write data to PI Data Archive?What are the main components of PI System architecture?How do I optimize PI Data Archive performance?How do I set up the PI Interface for OPC DA?How do I use PI Connector for UFL to import file data?What are the security best practices for PI System?How do I migrate PI Data Archive to a new server?How do I connect Power BI to PI System?How do I manage units of measure in PI System?What's the best way to learn PI System?How does PI time syntax work?What are good PI tag naming conventions?How do I access PI System data from Java or Linux?How do I set up PI AF notification rules?How do I use batch requests in PI Web API?How do I connect to PI AF Server using AF SDK in C#?How do I write VBA macros in PI ProcessBook?How does PI data compression work?How do I get real-time streaming updates from PI Web API?How do I troubleshoot PI AF Analysis errors?How do I write SQL queries against PI using PI OLEDB?What is PI interface buffering and how do I configure it?How do I search and query PI Event Frames?What is PI Integrator for Business Analytics?How do I fix the 'Point does not exist' error in PI System?What is the difference between PI ACE and AF Analytics?How do I manually enter data into PI Data Archive?What are AF hierarchy design best practices?What is a WebID in PI Web API and how does it work?How do I configure Kerberos authentication for PI System?Can I deploy PI System in the cloud?How do I calculate summaries (average, total, min, max) in PI?What are PI digital states and how do I use them?How do I connect PI System data to Grafana?How do I backup and restore PI Data Archive?How do I create custom symbols in PI Vision?How do I automate PI System tasks with PowerShell?What is AVEVA Connect and how does it relate to PI System?How do I check and handle data quality in PI System?How do I handle pagination in PI Web API responses?How does AVEVA PI System licensing work?How do I plan a PI System upgrade?