Build a Lustre High Performance Cluster

Coming Soon using v2.12.5 Overview Building a (Non HA) Cluster Todo – TCP LNET Overview Metadata Server Todo – Installing the Components – e2fsprogs, kmod, lustre Todo – Recommended Storage Needs (RAID-10 stripes) Todo – /etc/fstab entry for mount on boot. Object Storage Servers Todo – Installing the Components – e2fsprogs, kmod, lustre Todo – …

Infiniband Fat Trees

Overview Infiniband is primarily used in High Performance Computing (HPC) and provides a very fast network interconnect with an incredibly small latency. The Infiniband Iterface cards also implement Remote Direct Memory Access (RDMA) which allows the cards to place data directly into the memory of the end node so that no Kernel intervention is required. …

HPC Re-Engineering

Overview We have been tasked with proposing how we would re-engineer an aging SGI High Performance Computing platform. The platform is a 35 Node cluster running Red Hat 6.7 on each node with a Lustre file system mount and a /home file system presented via NFS from a Hierarchical Storage Node (over an FDR Infiniband …

LACP Implementation

Modern servers usually have multiple ethernet interfaces, and by combining them into an LACP trunk, you can pass VLAN’s over then to extend the amount of traffic over a bonded network link. Converting a Linux CENTOS server with 2 or more Ethernet interfaces to an LACP bonded network link couldn’t be simpler. The process is well documented …

Cloning to smaller Disks

A spate of failures of Samsung 850PRO SSD drives at a client’s site was the inspiration for this post. The SSD’s were used to boot the OS and were in a RAID 1 mirror when the second SSD had failed. Replacing the Samsung SSD would have been easy as its a hardware Raid, but the client …

Core Files in CENTOS 7

Since Centos 7 came out there has been a significantly growth in the number of core files appearing on machines I look after. Since these machines are all package built and maintained with YUM from the CENTOSrepositories then that is a bit of a worrying trend. As an aside, I’ve also noticed the number of package …

Repairing CENTOS 7 Journal Corruption

Most Linux distributions use to use flat text files to store logging information of both user and system processes prior to systemd being introduced. For decades this has worked quite well, however with the introduction of Systemd in the Red Hat/CENTOS 7 distribution, the logging of system processes is primary logged via the systemd-journald service. …

Striping Logical Volumes

A Logical Volume (LV) is created from a Logical Volume Manager (LVM) Volume Group (VG). The VG is the central storage component in an LVM based system, a VG can contain one or more Physical Volumes (PV), which are basically disks presented to the files system, partitioned as an LVM disk and then converted to …