Heavy system load by pagedaemon

[Iasen Kostov]

On Wed, 2006-05-10 at 15:43 +0300, Iasen Kostov wrote:
> On Wed, 2006-05-10 at 15:32 +0300, Iasen Kostov wrote:
> > On Wed, 2006-05-10 at 15:28 +0300, Iasen Kostov wrote:
> > > 	Hello
> > > I (probably) have some problem with a hosting server machine halting
> > > about once a 30 to minute (probably when the free memory falls under
> > > some threshold). At that moment the machine totally halts for about 5 sec
> > > then resumes normal work with load average around 9-15 (depends on
> > > system load) from about 2-3. The system is:
> > > (nice try :()
> >  FreeBSD 6.1-PRERELEASE #7: Wed Mar 15 17:35:21 EET 2006
> > 
> >  ACPI APIC Table: <PTLTD          APIC  >
> >  Timecounter "i8254" frequency 1193182 Hz quality 0
> >  CPU: Dual Core AMD Opteron(tm) Processor 270 (2004.56-MHz K8-class CPU)
> >    Origin = "AuthenticAMD"  Id = 0x20f12  Stepping = 2
> > 
> > Features=0x178bfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CLFLUSH,MMX,FXSR,SSE,SSE2,HTT>
> >    Features2=0x1<SSE3>
> >    AMD Features=0xe2500800<SYSCALL,NX,MMX+,<b25>,LM,3DNow+,3DNow>
> >  real memory  = 8724152320 (8320 MB)
> >  avail memory = 8315379712 (7930 MB)
> >  FreeBSD/SMP: Multiprocessor System Detected: 4 CPUs
> > 
> 	Hum, I was probably wrong about the free memory threshold - it
> currently has ~1600MB of free and it is doing it again. When the free
> memory falls under ~30MB it is just doing it more frequently.
> 
	And another odd thing (to me atleast):

#:> vmstat -s | grep "daemon\|fault"
        0 page daemon wakeups
        0 pages examined by the page daemon
  4992608 copy-on-write faults
    29021 copy-on-write optimized faults
    75167 intransit blocking page faults
 66956262 total VM faults taken
        0 pages freed by daemon

Acording to vmstat pagedaemon has never been awake. But after every 5
sec. lock I can see it eating ~25% of cpu (100/4 cpus) and I can see
processes in "*Giant" and "*vm ob" states. Worst thing it that the
machine is freshly rebooted and it was doing it even with ~6G of free
RAM and no swapping at all.

Kamal R. Prasad suggested that "Belady's anomaly" might occur. If it's
true or possible what could be done about this ?