Re: Monit memory usage

[Yogi]

Thanks for the reply.

> This is very likely related to the resource monitoring code in monit
> on Mac OS X. Resource monitoring on Mac OS X has been shaky. From
> your own mail, sent to the list a while ago:
>
> "It looks like Apple is phasing out support for /dev/mem and the
> Intel version of Tiger does not have it by default. It can however
> be enabled using a kmem=1 boot-arg. So I was wondering if this is
> the only workaround to get resource monitoring working or is there
> another way?"

Actually, this is happening on a PowerPC based XServe (on which I did
not have to use the above hack).

> If you remove the resource testing code from .monitrc the leak should
> go away. I get no memory leaks on mine iMac x86 in that case. I
> haven't tested for leaks using the hack above.

I removed all resource testing code, and it still adds 500K on every cycle.

> Is that the new intel based server or the old based on powerpc?

PowerPC

> If you feel adventures and have developer tools installed you can try
> to track down the leak using "/Developer/Applications/Performance\
> Tools/MallocDebug" Remember to start monit with the -Iv arguments as
> shown in the screenshot. But as I said I'm quite certain that this
> leak is either in our resource code or in apple's kvm library
> interface. Since the code seem to work okay on PPC  it may be that
> the hack above is the cause and the problem is in the kvm_* api.

Unfortunately I don't have access to MallocDebug on this server.

> In any case, I think we need to rewrite the resource code for Mac OS
> X and I plan to start by investigate "top" on darwin to see how apple
> grabs memory information.
>
>   http://www.opensource.apple.com/darwinsource/10.4.8.x86/top-17.4/

Thanks,
- Yogi