CARP setup : CPU load

nrgyz · Jan 4, 2008, 11:09 PM

Hi,

We setup CARP+VIPs one month ago and it is working great. The only concern I have is that CPU load is 30% steady. Before implementing, it was only 2-3%. Could it be considerated as normal? Anyone could confirm?

Thanks in advance.

sullrich · Jan 4, 2008, 11:39 PM

You did not setup carp sync on both hosts did you? Only firewall 1 should be syncing to firewall 2. #2 should NOT sync back to #1.

nrgyz · Jan 6, 2008, 1:29 AM

Sync is enabled (checked) on both master and slave. Isn't correct? I've followed the tutorial…

nrgyz · Jan 6, 2008, 1:32 AM

I forgot to say that only the CARP master is having the 30% CPU issue… The slave is still running at 3%.

sullrich · Jan 6, 2008, 6:21 PM

Take a screenshot of each of the hosts CARP setting screens.

nrgyz · Jan 7, 2008, 7:39 PM

Thanks for your quick reply. It is really appreciated!

Screenshot of the CARP master :
http://img134.imageshack.us/my.php?image=carpmasterec3.jpg

Screenshot of the CARP slave :
http://img230.imageshack.us/my.php?image=carpslavenv2.jpg

nrgyz · Jan 7, 2008, 8:33 PM

I can confirm with :

# tcpdump -i fwe0

that only CARP master is syncing to slave (not reverse)

sullrich · Jan 7, 2008, 10:09 PM

Your setup looks correct. Not sure why you would be seeing this added load. My CARP hosts are not showing this.

Can you try running a few commands:

top -I
vmstat -i
vmstat -s

nrgyz · Jan 8, 2008, 2:57 PM

# top -I
last pid: 28232; load averages: 0.66, 0.34, 0.27 up 53+12:19:21 09:15:58
42 processes: 1 running, 40 sleeping, 1 zombie
CPU states: 0.4% user, 0.0% nice, 0.0% system, 2.3% interrupt, 97.3% idle
Mem: 77M Active, 17M Inact, 51M Wired, 38M Buf, 286M Free
Swap: 1024M Total, 1024M Free

PID USERNAME THR PRI NICE SIZE RES STATE TIME WCPU COMMAND
5756 root 1 8 20 6588K 6096K wait 579:03 1.86% sh

# vmstat -i
interrupt total rate
irq0: clk 330038995 71
irq1: atkbd0 71 0
irq3: sio1 1 0
irq4: sio0 1 0
irq8: rtc 591744555 127
irq10: fwohci0 re* 2122065738 458
irq11: re0 atapci0 134476563 29
irq14: ata0 6560429 1
Total 3184886353 688

# vmstat -s
3996661449 cpu context switches
3086919878 device interrupts
159397933 software interrupts
2324797307 traps
1958236698 system calls
39 kernel threads created
10128798 fork() calls
484718 vfork() calls
0 rfork() calls
0 swap pager pageins
0 swap pager pages paged in
0 swap pager pageouts
0 swap pager pages paged out
469 vnode pager pageins
3167 vnode pager pages paged in
2003234 vnode pager pageouts
2034058 vnode pager pages paged out
0 page daemon wakeups
0 pages examined by the page daemon
265 pages reactivated
1701373918 copy-on-write faults
885999 copy-on-write optimized faults
415813077 zero fill pages zeroed
414164367 zero fill pages prezeroed
2874 intransit blocking page faults
2329291089 total VM faults taken
0 pages affected by kernel thread creation
2926195887 pages affected by fork()
255613759 pages affected by vfork()
0 pages affected by rfork()
2179053141 pages freed
0 pages freed by daemon
1936954364 pages freed by exiting processes
19535 pages active
5476 pages inactive
0 pages in VM cache
12961 pages wired down
72043 pages free
4096 bytes per page
774325484 total name lookups
cache hits (89% pos + 9% neg) system 0% per-directory
deletions 0%, falsehits 0%, toolong 0%

dotdash · Jan 8, 2008, 9:27 PM

It may be the firewire interface. My experience is that firewire is usable for a sync interface, but it hasn't been 100% stable…

sullrich · Jan 8, 2008, 9:36 PM

@dotdash:

It may be the firewire interface. My experience is that firewire is usable for a sync interface, but it hasn't been 100% stable…

Yep, that appears to be the culprit indeed.

If you are not using firewire, try turning it off in the BIOS.

nrgyz · Jan 9, 2008, 4:22 PM

Unfortunately I'm using the 1394 int. for the CARP sync. I thought it was stable…