The Qlustar HA Stack

The Qlustar HA Stack contains all components needed, to configure cluster head-nodes or storage nodes in a high-availability (HA) setup and protect them from SPOFs (single point of failures). It is based on the corosync + pacemaker infrastructure.

The Q-Leap team responsible for the Qlustar HA Stack has two decades of experience in designing, setting up and operating HA clusters for various types of mission-critical customer systems in industry, research and academia. Correctly configured clusters running the Qlustar HA Stack feature exceptional service availability achieved by efficient monitoring of the services they provide, combined with reliable and fast service fail-over in case of hard- or software failures. Contact us, if you need assistance in setting up or operate a reliable HA cluster.

Qlustar Resource Agents

A Pacemaker resource agent is an external program (mostly a shell script), that abstracts the service it provides. It presents a consistent interface for starting, stopping and monitoring the service. Such an agent is required for each type of service that is supposed to operate in HA mode. The Qlustar HA stack comes with a large set of well-tested resource agents. They cover the services needed on typical HPC, storage or cloud clusters. Among them are agents for:

  • ZFS pools and filesystems.
  • Lustre OSTs, MDTs, MGTs.
  • BeeGFS OSTs, MDTs, and management server.
  • Slurm control and database daemon.
  • NFS kernel server and its exports.
  • Samba file server.
  • Apache web server.
  • MySQL/MariaDB databases.
  • Oracle databases and listener.
  • MD software RAID devices.
  • LVM logical volumes.
  • Virtual IP addresses.

Using these resource agents within the Pacemaker framework allows to set up an HA cluster with any combination of the corresponding services as so-called resources. The configured services (resources) will then be protected from hard- and software failures.