Flashes (Alerts)
Abstract
Using ppc64 cpu to change the SMT mode of an IBM POWER8 Server may cause a system I/O hang if done concurrent with a high I/O load to an NVMe storage adapter or device.
Content
- Red Hat Enterprise Linux 7.2, little endian
- Red Hat Enterprise Linux 7.2, big endian
- Ubuntu Server 16.04
IBM Systems affected
- 8247-21L
- 8247-22L
- 8284-22A
- 8335-GCA
- 9119-MHE
- 9119-MME
- 8247-42L
- 8286-41A
- 8286-42A
- 8001-12C
- 8001-22C
Affected Hardware
- EC54 PCIe3 LP 1.6TB NVMe Flash Adapter
- EC56 PCIe3 LP 3.2TB NVMe Flash Adapter
- EC55 PCIe3 1.6TB NVMe Flash Adapter
- EC57 PCIe3 3.2TB NVMe Flash Adapter
- EKAE PCIe3 Dual Port NVMe Host Bus Adapter
- EKEE PCIe Dual Port NVMe Host Bus Adapter
- EKN2 PCIe3 1.6TB NVMe Flash Adapter
- EKNA 800GB NVMe 2.5-inch SSD
- EKNB 1.2TB NVMe 2.5-inch SSD
- EKNC 1.6TB NVMe 2.5-inch SSD
- EKND 2.0TB NVMe 2.5-inch SSD
- EKNJ 800GB NVMe 2.5-inch SSD
- EKNN 3.2TB NVMe 2.5-inch SSD
Symptoms
Without a fix for this issue, if the SMT mode is changed using ppc64_cpu, a system I/O hang may occur, requiring a reboot of the LPAR or system to recover.
It is recommended to quiesce I/O operations to all NVMe storage prior to changing the SMT mode in order to avoid this exposure.
IBM is working closely with Red Hat and Canonical to make fixes for this issue available through the normal fix release process. This alert will be updated once an outlook is available for a fix of this issue.
Brian 2016-06-23
Was this topic helpful?
Document Information
Modified date:
25 September 2022
UID
isg3T1023933