Notifications

Objectives

• Get alerted when things go bad.
• Don't depend on having to watch the machine 24/7
• Integrate with the journal
• Easy accessible

Tentative Design

• Notification area
• Every message goes through the journal. Only the ones of a certain level are consithered important enough to trigger a notification event (inside cockpit or via email/xmpp).
• This however requires that by default a system do not spit out a lot of stuff by default to the journal on that level.

How to alert

• Email and XMPP

What to alert for

• Too high CPU/Memory/Disk IO
• Dying disks
• Crashing services?
• Power failure?
• Failing Internet connection?
• Packet loss?

Relevant Art

http://cdn.arstechnica.net/wp-content/uploads/2013/11/Screen-Shot-2013-10-25-at-2.57.59-AM.png

Comments

• Multi-machine vs single machine notifications needs to be discussed and figured out.

See Also