My experience in system administration based mostly on Linux systems and started in 1993
in Joint Institute for Nuclear Reasearch
(JINR), Dubna, Russia.
My first project was a setup of small Linux cluster for HEP group. As time progress,
I've been carry out many system administration duties.
Among them:
- Linux clusters setup and maintenance
- DB administration (primarily MySQL)
- kernel maintenance for Clue-d0 cluster
- software deployment and cross-platform ports (OSF-Solaris-Linux)
- system installation, troubleshooting, configuration,
The most interesting project was based at
FNAL where we setup a large Linux cluster
based on user desktops.
Cluster
lin
ux
environment for
D0 experiment
(
Clue-d0) was born as an idea to use
idle CPU(s) of user desktops for analysis purpose. The HEP experiments is always a
collaboration of different Institution, Universities and labs who brings their
hardware to fulfill a goal of the project. That leads to heterogeneous hardware
configuration of the custer. Based on our experience we
recommended
some hardware vendors for our collaborators. The underlying OS was always
RedHat-based distributions (including different version of RedHat and Fedora Core). The
user accounts were managed throw LDAP server where each group had own LDAP slave
we kept in sync with a master. To accomodate various hardware as well as D0 requirements
e.g. XFS as underlying file system for RAID file servers, we customized the
underlying kernel for our needs. All file servers were equipped with Gigabit NIC.
The underlying OS patches were provided in a form of RPMs and distributed in a cluster
through autorpm mechanism. All users were able to run their tasks in interactive and
batch queues. By the time I left the cluster consists of 250 nodes
and was proven to be reliable alternative to physics analysis farm.