Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe

HK: Fachverband Physik der Hadronen und Kerne

HK 70: Instrumentierung XIV

HK 70.7: Vortrag

Freitag, 19. März 2010, 15:30–15:45, HG ÜR 6

Introducing high availability to non high available designed applications — •Pierre Zelnicek¹, Oystein Senneset Haaland², Udo Kebschull¹, and Volker Lindenstruth³ — ¹Kirchhoff Institute of Physics, Ruprecht-Karls-University Heidelberg, Heidelberg, Germany — ²Physic Institut, University of Bergen , Bergen, Norway — ³Frankfurt Institut für Advanced Studies, University Frankfurt, Frankfurt, Germany

A common problem in scientific computing environments and compute clusters today, is how to apply high availability to legacy applications. These applications are becoming more and more a problem in increasingly complex environments and with business grade availability constraints that requires 24x7x365 hours of operation. For a majority of applications, redesign is not an option. Either because of being closed source or the effort involved would be just as great as re-writing the application from scratch. Neither is letting normal operators restart and reconfigure the applications on backup nodes a solution. In addition to the possibility of mistakes from non-experts and the cost of keeping personnel at work 24/7, these kind of operations would require administrator privileges within the compute environment and would therefore be a security risk. Therefore, these legacy applications have to be monitored and if a failure occurs autonomously migrated to a working node. The pacemaker framework is designed for both tasks and ensures the availability of the legacy applications. Distributed redundant block devices are used for fault tolerant distributed data storage. The result is an Availability Environment Classification 2 (AEC-2).