Times of System Maintenance

Regular Intervals

Each Thursday, from 07:00-08:00.

These hours are not explicitly announced. They are defined to make outages predictable and expected for our users, so the non-availability of services can be dealt with constructively beforehand.

Irregular Intervals

In addition, it can be expected, that a system has to be put into irregular maintenance twice a year for four hours.

Irregular maintenance will be announced.

Repairs

Systems fail spontaneously and have to be repaired. The usual service contract with vendors is next business day on site support. In exceptional circumstances, incidents happen on fridays or a first fix is not successful. 48 hours for incident response per year should be taken into account.

Effects

During maintenance hours, server operation is unreliable. You should expect unavailable servers and sudden downtimes of services. During the rest of the day, performance may be degraded due to continuing background processes.

Misconceptions

• Misconception: nothing IT-related will work during the maintenance interval. That is not correct. In fact, during that hour, only some systems become unavailable, but not all, if any.

• Misconception: outage of services is a full hour. That is not correct. Effect for most system is only a few minutes. Due to the large storage and multitude of services running on the system, the admin groupware server ("Outlook/Exchange") takes longest with about 20 minutes.

• Misconception: all systems are affected. That is not correct. Systems will only be affected, if there are outstanding patches required to be installed to enhance security.

• Misconception: systems are offline in each maintenance hour. That is not correct. There are also maintenance intervalls without any required action on the IT/Tk infrastructure.

Necessity of System Maintenance

Here are examples of incidents around the world which underline the importance of having up-to date systems:

• Datenbank der Universität von Kalifornien gehackt
• (to be continued)

Workarounds

File Servers

The various file servers, addressed as e.g. \\admin01\..., \\admin02\..., \\faculty01\..., \\faculty02\..., etc., sometimes mapped as P: drive in Windows clients, will become unavailable. The workaround is to copy the file(s) to be manipulated to a local hard disk, edit the file(s) locally, and copy them back after the maintenance hour.

Or, if the files have been open from the file server before the server became unavailable, either wait until the server is back online and continue as if nothing happened, or use the "File -> Save As..." dialogue for creating a local copy and move that to the file server after it is online again.

Printing

Jacobs University print services are also run by Windows servers and experience the same effect. Locally attached desktop printers can be used as a workaround. Admin users can use the faculty print server during downtime of admin servers.

Login and Roaming Profile

If logging in throws a message due to an inaccessible roaming profile, one can still log on with a local profile. Changes to documents within the local profile will not be synchronized to the roaming profile server until next logon with connection to the server.

Admin Groupware

Outlook clients will loose its connection to the admin Exchange groupware server for Admin Email. After a timeout, Outlook can be used in offline mode operating on the information stored locally, including calendar, contacts and emails. Emails can still be searched and, and new emails written which will be stored in a client-local Outbox, until the Exchange server is back online.

After Outlook having lost connection to the Exchange server, it is advisable to close Outlook and start it anew after the Exchange server became available again to circumvent reconnection issues with some Outlook versions. Also make sure, that the "Online" mode indication in the lower right edge of the Outlook window appears.

For non-admin emails, the web interface of Jacobs University Email can be used.

Various Application Servers

Workarounds depend on the application and server. Please negotiate a Service Level Agreement for the server. Existing SLAs can be seen at All SLAs.

Internet, Firewalls, Network

There is no real workaround except copying information locally and removing it afterwards. On the bright side, outages are usually very short and rather seldom.

Availability Target Calculation

Please see the attached Excel spreadsheet for the calculation.

Business hours are assumed to be 8am to 6pm Mon to Fri for this calculation, minus "Betriebsurlaub" of estimated 2.5 days per year.