Server load monitoring

[snowcreative]

Here's what just happened, and it is something that happens somewhat frequently.

The server one of my clients is on (Rhea) had a load spike up to 200 at 1:50pm. It came down quickly, but then went up to 222 at 2:25, then came down again, then back up to 325 at 2:50 and kept going up. It reached 510 at 3:10, and HTTP crashed on the server. I sent in a support request at 1:50, hoping someone would see it in time to respond and prevent a crash. That didn't happen; 20 minutes later, at the time of the crash, the ticket still hadn't been assigned to anyone. So, from the time of the initial spike, it took and hour and twenty minutes before the crash happened.

Here's the question. Shouldn't there be some sort of load monitoring on your servers to alert techs to this sort of thing? In the past, the techs have said "thanks for letting us know" when I report these load spikes, but why should I have to report them at all? I have a script on my server that checks the load periodically and sends me an email when it goes well above the normal range. If the load stays up, or goes down but goes up again, I submit a support ticket. Why isn't something like this in place on all the servers to alert the techs? That would prevent some of these server crashes from happening.

[JPC-Masood]

We have monitoring script which reports this and logs it but that does not resolve the issue immediately as it can take time and sometime days to pinpoint which client's script is causing the problem. Thank you for opening the ticket.

[snowcreative]

Does that mean the techs start investigating right away when there are load spikes? Their replies always sound like they were unaware of the spikes that I report.

[JPC-Masood]

Level 1 techs do not work on server load issues and all your tickets go to level 1 first. Level 3 techs are the ones who work on them and yes this is what they do primarily other than working on escalated tickets. They are working on load issues and other server issues all the time as soon as first sign of it comes through different monitoring tools.

[snowcreative]

Ah, I see! Thanks for the clarification. And, good to know someone's eye is on this type of thing.

[Bell thorpe]

If you're going to quote someone else's material, should you not at least provide an attribution?

Or better yet, just a url so those interested can click and read?

[Vin DSL]

Following are few commands used for the Linux server load monitoring and maintenance

Commands:

top: - used for the viewing current process going on the server. You will get a dynamic table of running process as a output. [...]

Gotta say...

I'm rather partial to HTOP!

[yorch]

Agree, htop is just great.

And now the OT:
Hey Vin, really like your desktop.. which theme are you using?? also, is that conky right?
Thanks!