HPC - In-house Cluster for High-performance Parallel Computing

Guides

• User Guide
• Admin Guide
• Setup Guide
• Remote Server Management
• Hardware Specs

Cluster Configuration

Service Logs

• 2022/04/25 - Node 1 faulty memory replacement

  Two faulty memory sticks from slots CPU1_DIMM_C1 (not sure; maybe B1) and CPU2_DIMM_B1 have been replaced. In addition, CPU1_DIMM_A1 has also been replaced while diagnosing the faulty memory slot location.

• 2022/12/19 - Node 8 faulty memory replacement

  System Event Log:

  EventID:0074    Time:Wed Nov 30 20:57:31 2022    Controller:SMI Handler    SensorType:Memory    SensorName:Mem err Sensor    Description: Post Package Repair Runtime Request. Rank: 0 CPU: 1 DIMM: E1. - OEM  - Asserted
  

  One (or multiple?) faulty memory replaced.

• 2023/05/10 - Node 6 memory ECC error (E1)

  System Event Log:

  EventID:0587    Time:Sat Apr 29 18:17:27 2023    Controller:SMI Handler    SensorType:Memory    SensorName:Mmry ECC Sensor    Description: CPU: 1, DIMM: E1 DIMM Rank: 1. - Correctable ECC / other correctable memory error - Asserted
  EventID:0588    Time:Sun Apr 30 17:50:19 2023    Controller:SMI Handler    SensorType:Memory    SensorName:Mmry ECC Sensor    Description: CPU: 1, DIMM: E1 DIMM Rank: 0. - Correctable ECC / other correctable memory error - Asserted
  

  One faulty memory from CPU1_DIMM_E1 has been replaced.

• 2023/05/10 - Node 9 Omni-Path PCIe speed issue

  It was identified that the Omni-Path NIC only used 1x lane out of 16x, resulting in a reduction in MPI communication speed. The solution was to move the NIC card to another riser slot. It was unclear as to why the first riser slot only supports 1x speed.