Services Unit Forum Deployment

In all that follows, the assumption is made that Appleton Tower and the Forum are in too close proximity to act as credible offsite facilities for each other. This document therefore deals with AT and the Forum as one location - central - with actual allocation of servers to the two central machine rooms to be determined later. For an initial pass at planning server location and requirements, see ServicesUnitForumPlanningServerDisposition.

Definitions

  • Disaster recovery priority - The priority given to restoring the service in the event of its loss.
  • Replicated data? - Is the data that the service relies on replicated/mirrored offsite?

Tape Backup

Disaster recovery priority - low (it is assumed that high priority data will be recovered from mirrors which are dealt with separately)

Replicated data? - yes, NSR indexes

Main Hardware location - backup server (one of the sun file servers) and tape unit located centrally.

Backup Hardware provision - Backup server (another Sun file server) and backup tape unit (probably single drive) located at KB. Fire safes in both locations with weekly rotation of tapes to KB.

Notes - This assumes a tape backup regime similar to the current one. This may have changed by time of the move to the forum.

To be clarified - Will the school van still be available for the movement of tapes?

Bugzilla

Disaster recovery priority - low

Replicated data? - yes, bug database

Hardware location - central

Backup hardware provision - no dedicated provision

Corpora

Disaster recovery priority - medium/low

Replicated data? - some, part of standard mirroring service.

Hardware location - central, possibly ECDF

Backup hardware provision - no dedicated provision

CVS/SVN

disaster recovery priority - low

Replicated data? - yes, repository contents

Hardware location - KB

Backup hardware provision - no dedicated provision

External Access gateway

Disaster recovery priority - high

Replicated data? - no

Hardware location - Ideally one server at each of three main sites. At minimum, one central server and one server at KB.

Backup hardware provision - servers will substitute for each other as required.

File/Mirror services

Disaster recovery priority - high

Replicated data? - some

Hardware location - roughly equal provision of servers and fibre attached storage at both central and KB sites

Backup hardware provision - as above

Notes: Allocation of master and mirrored data to be split equally between central and KB site. File servers will act as mirror servers during quiet periods. The current collection of 18 mirror servers will be condensed into a much smaller number of more powerful fibre attached machines.

FTP responder

Disaster recovery priority - low

Replicated data? - no

Hardware location - currently at KB shared with DCS legacy web

Backup hardware provision - no dedicated provision

Jabber

Disaster recovery priority - high

Replicated data? - no

Hardware location? - central

Backup hardware provision - no dedicated provision

Legacy FTP

Disaster recovery priority - low

Replicated data? - yes, contents of FTP site

Hardware location - central (shared with legacy web server)

Backup hardware provision - no dedicated provision

Mail

Disaster recovery priority - high

Replicated data? - yes, user mailboxes, mailing list archives

Hardware location - main mail/list server central. Virtual mail relay/Authenticated SMTP server KB

Backup hardware provision - machines at either site will substitute for each other as required. this may require a respec of the smtp machine.

Notes - by the time we are in the forum the vast majority of commodity mail services should be being provided by central staffmail service reducing load on main mail server and making substitution of services feasible.

Printing

Disaster recovery priority - high?

Replicated data? - yes, printcaps, ppd files

Hardware location - For each floor of forum - one HP8XXX class BW printer in south printing room, one HP4XXX BW and one Canon 2880 colour MFD in north printing room, high capacity printer/photocopier in school office. Low capacity BW printer at KB ( in the machines room?. Two print servers in Forum, one in AT, one at KB. Continue having one printer wire for school available at all three sites.

Backup hardware provision - servers will be able to serve printer at any site.

Room Booking Service

Disaster recovery priority - low

Replicated data? - yes, room bookings

Hardware location - central, currently shared with jabber service

Backup hardware provision - no dedicated provision

Notes - this service may well have transitioned to the central ediary system by the time of the move to the forum

Samba

Disaster recovery priority - high

Replicated data? - yes, some part of general file system, some specific to admin samba server

Hardware location - admin server central, commodity server KB

Backup hardware provision - each server will stand in for the other

Notes - this service will gradually wither as data transitions to AFS

Web

Disaster recovery priority - high

Replicated data - yes, contents of web sites

Hardware location - main web server, homepages server (shared with wiki) central. Legacy web server, group webserver, commodity cms server KB.

Backup hardware provision - procedures will be put in place to allow serving of at least frozen copy of websites at other site.

Notes - Some groups and legacy web servers rely on NFS mounted data, it's not all local to the web server hardware.

Wiki

Disaster recovery priority - high

Replicated data - yes, wiki contents

Hardware location - central (shared with homepages server)

Backup hardware provision - Web servers at KB, probably commodity cms server

Notes

For the anticipated disaster recovery procedures to work, there is a need for a high degree of interconnectivity between the servers and fibre channel storage at each site. It is assumed that even in cases where the master data is currently held on local disk for performance reasons, it will be acceptable to hold the replicated data on fibre attached storage and accept any shortfall in performance that may occur as a result of this should disaster recovery procedures need to be put in place. It would be prudent to locate at least one general purpose fibre attached server at KB which could be used to provide services such as jabber should the need arise.

-- CraigStrachan - 04 Oct 2007

Topic revision: r4 - 01 Nov 2007 - 12:43:47 - CraigStrachan
 
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback
This Wiki uses Cookies