Enterprise Storage, Maintenance, and Outages

Enterprise storage is a collective service used by many of our servers to accomplish tasks requiring large amounts of storage, disaster recovery mirroring, storage snapshots, and extremely fast disk. It is also the continuing policy of DoIT that local storage not be used unless deemed absolutely necessary, and then only with the permission of SEO management. This is done for enterprise disaster recovery purposes.

The enterprise storage infrastructure consists of a storage area network (SAN) as well as a number of disk arrays. Maintenance, both proactive and reactive, on the enterprise storage infrastructure is just as important as on a host in order to ensure reliability and availability. Because storage is so important to a host there is quite a bit of reliability built into the infrastructure, so much maintenance can be accomplished without disrupting normal services.

The enterprise storage team supplies the recommended fibre channel HBAs and multipath software for vendors supported host configurations. Hosts that are unable to meet the enterprise storage compatibility demands will not be connected to the storage infrastructure or supported. The enterprise storage vendors offer a mechanism to develop support for unsupported OS and hardware combinations, and the Linux team will submit the configuration for review.

From time to time occassions arise where the enterprise storage team will recommend that the hosts be brought offline for the maintenance, or that special procedures take place. During these maintenance periods the POST Linux team will comply fully with the recommendations of the enterprise storage team and either follow the procedures, take the hosts offline, or both. Hosts which are in compliance with DoIT's local storage policies and employ both local and enterprise storage for applications will be brought offline as well. Though an application may only rely on local storage, separating the applications from each other, and the dependencies such applications might have, is difficult and time consuming. If a host employs enterprise storage it is classified and treated as such, and if it does not, it does not; there is no intermediate classification for partial treatment of the host based on whether an application employs local storage or not.