Skipdawg

Well I am going to assume this happened again about 12 midnight Florida time this Saturday. I missed it so could not get a snap shot of it this time. For all looks good now.

But I heard from 3 members that for about 15 minutes the forums were just barely crawling.

I'm really starting to wonder? Support can't do anything without seeing it. And I can't do anything about it so what to do?

__________________

Skipdawg

Just a FYI update. Things so far seem to have simmered down. So far.
With my forums I moved all email activity with webmaster functions from domain email to GMail. See how that affects some of this mystery activity.

Also I learned some stuff about server load and know not to rely 100% on what shows in cPanel and WHM as a high server load. That is just showing what is going on for the moment.

I was told by a friend who works with servers with his company to watch the server load by the 1, 5, and 15 minute server load intervals and if the load shows high at the 5 and or 15 then to contact support and get them on it.

And I was also told anything could cause a 100% CPU spike even if the forums or other things I have in pace are inactive. Activity in MySQL or Apache updating or such can do it too and not to worry about it until it is a persistent 80% or better for like over 5 minutes at a time.

So hopefully this will help others too.

__________________

Skipdawg

Well I just caught it acting up again. Image attached for load at the moment. Forums were crawling like snails on ice.

The server load average: 6.69, 5.44, 3.49 glad I learned how to look for that.

Attached Thumbnails

 

__________________

Skipdawg

Well on it's way to Admin (3rd Level Support)
Ticket number for those who can and would like to followup is JEA-925641

I sure hope they can nail it down this time. I know folks with 10 times the forum size and activity who don't see half the resource use I do.

__________________

Skipdawg

Well 3ed level can't find a problem. Guess I have to deal with it till it becomes a pain in the butt and then if it dose just dump it all and find a new hobby.

__________________

cowboy

Briefly looking into the mystery of Server loads, CPU and memory reporting, I describe it to my beginner classes similar to the analogy of a city and it's street system.

The CPU is the "City Hall Dictator" which processes everything. The data coming in and going out travels on it's freeway system. A 32 bit processor, 32 lanes of traffic. A hyper-threading 32 bit processor, a double deck 32 lane interstate.

The memory is the "toll gate Sargent" dispatching data to a lane.

The Server Load (SL) is the traffic awaiting the Sargent's lane assignment. Server Loads are a total of all lanes on all decks and should be divided by the total decks to make sense. A SL of 4 on a 4-CPU system equates to 1.

A SL of 1 is equal to all lanes being full and no wait at the toll gate. A SL less than 1 means all data lanes are free and there is no waiting at the toll gate. A SL greater than 1 indicates all lanes are at capacity and there is a backup at the toll gate guesstimated on 1 as 100%. A SL of 2 suggesting it will take twice as long to get through the 1-CPU processor (SL 8 on 4-CPU).

On dedicated and shared servers all traffic lanes are available to the Sargent, however, he keeps a detailed log of each user. Generally there are only minor traffic backups. An example being many accounts with users in the same time zone and everyone using e-mail at the same time in the morning or after work.

When an account consistently uses the majority of the system and causes toll gate delays the server administration checks the Sargent's log, identifies them, and corrects the problem, like "road hog" suspension.

With VPS a different approach is taken. You are given a set amount of the toll gate lanes you can use, 10% is common but not set in stone. All of your lanes could be full and your on-ramp jammed, but, you are the only one waiting and show a very high SL and your neighbor has open lanes and a very low SL.

Programs or services set to run in the background are often overlooked as a cause. Photo albums sometimes trigger resizing after low server loads are reported.

A database backup of your site, for example, which took 1 minute when you were on a shared server and able to use 100% of the server, will take maybe 10 minutes to process on your 10% of the VPS and will raise your SL reporting figure substantially until it is done. This is the drawback of VPS when you are the one doing the backup, and your benefit when your neighbor is doing their backup.

Account backups, especially run as cron, get intensive as sites grow and tax the server for everyone. A polite cron script will check server load and only proceed if the server load is low.

Again I state this is an analogy I use and is not meant to be a technical description in any way.

__________________
Where would you be if you were at the highest court in the land (US)?

Skipdawg

cowboy that is a very kewl way to look at it. that should help anyone understand it all better.

I think something is asleep on the job in that though. For I've been told they don't see what I am seeing and showing them in the logs.


So that is telling me that what ever is use to check the logs is inadequate or the VPS was not setup right.

So if someone is not sleeping maybe they are AWOL hehehe

__________________

Skipdawg

Well I see activity behind the curtain uses some server load, CPU and memory.
In the VPS control panel under VPS services -> System processes I caught YUM active for a moment. But when I refreshed it was done.

Now I know I'm not and learning where to look for issues.

__________________

Skipdawg

Who can decipher this?

It was pushing the CPU to 100% and while this was going the server load was sitting at 7.97 (4 cpus)

Nothing else was going and no one was on the forums at the time.

__________________