Using IoT for server alerts

[aigini82]

Hi,

I have an idea of using IoT for servers to send alerts to warn us if there is a possibility of server crash due to any reason, eg hardware failure, etc.

This is to avoid the disaster in the first place. From my past experiences as a system admin working in other companies, there is no proper alert system on device/hardware failure in servers. We only know the server has a problem, eg smart array failure; once the server crashes. This is already too late, and the production system will be down for many hours which causes tons of revenue loss. I would like to avoid this at all costs.

How do I start my research/investigation to implement this idea? I am a novice in IoT.

Regards,
aigini82

 

[mintjulep]

Start with figuring out to know when something is about to fail.

 

[SamBones]

I don't know if this is a problem needing an IoT solution.

From my experience, there are warnings to possible upcoming hardware failures. Things like individual drives in an array, or an array controller. Maybe problems with a network card or memory sticks. These often show up as warnings or minor errors in the system log(s). That would be like the file /var/log/messages, 'dmesg' output, or the Windows event logs. Very often, hardware starts complaining here before an actual hard failure. These are all accessible from the operating system, so those are best monitored by something running on the machine. There are commercial products for this, and it's pretty easy to grow your own if you can find a way to catch all possible faults.

As far as detecting from outside the box, which is what IoT implies, you have two choices (as I see it). One is connected to the network and 'looking' that way, or completely disconnected and looking some other way. The problem with either of those is exactly what mintjulep says. You first have to figure out what you are monitoring to determine that something may fail.

Just looking at my servers, they have lights all over them. The lights can be either green, yellow, or red. When the server is happy, all lights are green. Some go yellow when it's having an issue. They go red if there is some kind of hard fault. And covering all conditions, they go out if there's a total failure of the system. So I could see a Raspberry Pi with a camera, tell it what a good all green condition looks like, and text me if it "sees" anything different. Maybe also a way to ping that Pi and see what it sees myself. That would automate my warning light check.

But mintjulep is completely correct. You first need to figure out what the IoT device would need to be monitoring to be able to predict an upcoming failure.