Memory error recovery method and device

Abstract

The application discloses a memory error recovery method and device, wherein the method comprises: adding up memory errors correctable; if the number of the memory errors correctable reaches a preset threshold, determining a physical address range affected by damaged memory particles causing the memory errors correctable; killing all processes using the physical address range, and releasing corresponding memory addresses; subjecting all pages comprising the memory addresses to offline processing; restarting all the killed processes. The method and device enable the memory particles already experiencing continuous correctable errors to be called again by an operating system, and accordingly more severe non-correctable errors due to the memory on the damaged particles are avoided to avoid downtime.

Description

technical field [0001] The invention belongs to the technical field of operating systems, and in particular relates to a memory error recovery method and device. Background technique [0002] With the promotion of big data-related industries, IT suppliers are using more and more large-scale servers to provide corresponding computing and storage capabilities. Taking the mainstream 4-socket server as an example, the maximum configuration is 96 pieces of memory. In the data center, the number of tens of thousands of sets makes the number of memory in the data center increase sharply. Along with the increase in the number, there is also the probability of memory errors. Memory errors are divided into correctable memory errors and uncorrectable memory errors. The memory itself has ECC error correction logic. A correctable error means that an error of 1 bit detected by the memory can be corrected. Errors of more than one bit are collectively referred to as uncorrectable errors. ....

AI Technical Summary

Problems solved by technology

[0003] Memory granules that have continuously occurred correctable errors are usually physically damaged. Damage to the memory on the granules will lead to more serious uncorrectable errors, and may even cause downtime problems. However, there is no relevant prevention in the prior art measure

Images