Insane CPU Usage

[cbk1994]

On Era Dev, we're constantly seeing things like this in our stats:

Originally Posted by /stats

Top 10 of npcs taken most CPU time:
1. 14.767989811 percent -System/Commands (commands_items,commands_def,commands_staff)

This is absolutely ridiculous; there is no feasible way the script could be this CPU-intensive. The only command used was a single summon. This happens to virtually all of our scripts, from the simplest to the most complex. I don't notice any pattern or specific time this happens, but at some points the server is so laggy we can't do anything, even on RC with no scripts running. I have checked out the scripts with function profiling and all other tools we have, but there is no reason for this to be occurring.

This is clearly a problem with the server and not our scripts, and I'd like to hear from Stefan about what we can to to fix this. It is impossible to get any kind of clue how efficient our scripts are when they're constantly reaching 15percent on the CPU usage list for no reason.

[WhiteDragon]

If you already did function profiling on everything and analyzed /stats when the problem was happening, I imagine this would be a problem Stefan would have to take care of.

It's possible there is something else on the machine making iowait go up and therefore artificially boosting all the "CPU usage" percentages on the GServer.

It could also be one of your scripts doing too much I/O. Is your memory usage too high? Perhaps it's swapping.

There are a number of things that could be happening, but I guess it'll be tough to tell without someone with access to check these things coming.

[Switch]

RELEASE THE CPU EFFICIENCY TOOL!/caps

[cbk1994]

Originally Posted by WhiteDragon

If you already did function profiling on everything and analyzed /stats when the problem was happening, I imagine this would be a problem Stefan would have to take care of.

It's possible there is something else on the machine making iowait go up and therefore artificially boosting all the "CPU usage" percentages on the GServer.

It could also be one of your scripts doing too much I/O. Is your memory usage too high? Perhaps it's swapping.

There are a number of things that could be happening, but I guess it'll be tough to tell without someone with access to check these things coming.

We're using approximately 8 MB (8.1937561) of memory, so that's not the problem.

As far as I know, virtually all UC servers are on the same machine, so I wonder if that's the problem? Perhaps we could be transferred to whatever machine Era is on.

[WhiteDragon]

Originally Posted by cbk1994

As far as I know, virtually all UC servers are on the same machine, so I wonder if that's the problem? Perhaps we could be transferred to whatever machine Era is on.

I think the best bet is that another machine is doing something intensive that makes your script wait on some resource and say that it's using more CPU than it actually is.

[cbk1994]

Originally Posted by WhiteDragon

I think the best bet is that another machine is doing something intensive that makes your script wait on some resource and say that it's using more CPU than it actually is.

I'm wondering if it has something to do with Stefan's backups as well. I know that has caused problems like this for servers in the past.

[xXziroXx]

We get the same once a day on Maloria around 00.30-02.00 GMT+1, scripts containing basically no script at all start complaining about 20% CPU usage while the server itself lags unbearably much.

Every day, same time.

[Pelikano]

You might want to show this to Crow, I think he once had the same problem with a script... nick-system?

[Cloven]

There's a massive lag issue at the same time daily on Delteria Dev as well. NC usually dies reliably at the same time along with it (or at the very least, does not work properly for a short period of time). In one instance I've actually had a script revert to an older version of itself without explanation. o_O

[cbk1994]

Originally Posted by Pelikano

You might want to show this to Crow, I think he once had the same problem with a script... nick-system?

That was clientside lag, and it had to do with him using player attributes to display nick names .



Sounds to me like this is a common issue, and perhaps Stefan is spreading the machines too thin?

[WhiteDragon]

Judging by what everyone is saying, it sounds like Stefan runs some sort of backup which somehow uses a resource (probably i/o) that makes all the GServers wait on stuff and artificially boosts their CPU usage.

I imagine something should be able to be done by it considering there is this sort of problem in any real-time system that requires backing up.

[Switch]

Originally Posted by cbk1994

Sounds to me like this is a common issue, and perhaps Stefan is spreading the machines too thin?

And/or maybe a server is doing some lag-intensive script at that time.

[cbk1994]

Originally Posted by Switch

And/or maybe a server is doing some lag-intensive script at that time.

Well, besides the fact that there is no such thing as a "lag-intensive" script, I think you'd be hard pressed to make a script capable of pretty much crashing every single server on the machine you're on.

[Crow]

Originally Posted by cbk1994

That was clientside lag, and it had to do with him using player attributes to display nick names .

This.

I'm thinking as well that the issue is related to the machine Era Dev is on.

[xXziroXx]

Originally Posted by Switch

And/or maybe a server is doing some lag-intensive script at that time.

Unlikely, since it happens every day at the same time.

Originally Posted by cbk1994

Well, besides the fact that there is no such thing as a "lag-intensive" script, I think you'd be hard pressed to make a script capable of pretty much crashing every single server on the machine you're on.

... I did it once.