Patenteted Technology - Protection against data loss
- Effective protection against split-brain and deathmatch
- Independent communication channel
- No admin-fencing necessary
- Extendable with Collax Fencing Expander
- Automatic configuration
The patented Collax Fencing Device ensures data integrity at any time and at any operating status.
Collax Fencing – effective protection against Split-brain and Deathmatch
Collax HA-solutions provide reliable protection against split-brain and deathmatch and effectively isolate an erratic node in order to protect the rest of the system from loss of data.
Data corruption can occur if:
- One Node failed or has unintentionally been switched off.
The remaining node takes over control as well as all functions of the failed node. High availability is guaranteed. Necessarily both nodes have to communicate with their counterpart.
· The interconnect has been interrupted.
As consequences, both nodes assume that the corresponding node is no longer available and try to take over the tasks and functions of the other node. Each node tries to control the shared storage resource. This may lead to data chaos and data lost. A subsequent combining of the data is hardly possible.
Which dilemma can occur?
Even if the node realizes that its counterpart node is no longer available, there is no way to clearly identify the condition “node-failure” or “interconnect-interruption”. Massive data loss could occur if the node assumes a wrong failure condition and starts the corresponding countermeasures.
The Collax Fencing Device ensures that the fully functional node switches off the defect node. Therefore, it can be clearly assessed that the remaining node has control and is allowed to take over all functions.
What differentiates our solutions from other vendors on the market?
All common procedures are based on a third control entity in order to solve the above mentioned dilemma. Please find below the possibilities with the associated disadvantages:
- Admin-Fencing: the IT-administrator manually rebuilds the high-availability of the nodes.
- Node-Quorum: this solution requires at least three nodes. There is still the risk of restricted availability or even Split-Brain.
- Quorum with two nodes: A two-node cluster requires a quorum on a computer that doesn’t belong to the cluster. Due to the deep technical know-how of the administrator the solution is cost-intensive. However, there is still the risk of unclear cases.
- Quorum by SAN: An external SAN can be used as quorum. Even in this case there is the danger of unclear cases.
- Fencing by Stonith: PDUs and Lights-Out-Devises must be configured. Deathmatch can still occur.
All these techniques have in common that they need additional equipment or an administrator to ensure high-availability for a two node cluster.