Re: miserable performance of 2.6.21 under network load
On Tue, Jul 24, 2007 at 09:13:47PM +0200, Andi Kleen wrote: > You have slab debugging enabled. That makes everything slow. > It costs you ~66% of your CPU time. Disable it. Past peak for the day, but this does look like the culprit. Thanks to all who replied. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
From: Aaron Porter <[EMAIL PROTECTED]> Date: Tue, 24 Jul 2007 11:49:09 -0700 > samples %app name symbol name > 914379 48.8404 vmlinux-2.6.21.5 check_poison_obj > 341920 18.2632 vmlinux-2.6.21.5 poison_obj > 37355 1.9953 nf_conntrack (no symbols) You have SLAB debugging enabled, turn it off if you want decent performance. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
Aaron Porter wrote: On Tue, Jul 24, 2007 at 08:48:00PM +0200, Andi Kleen wrote: Aaron Porter <[EMAIL PROTECTED]> writes: I'm in the process up upgrading a pool of apache servers from 2.6.17.8 to 2.6.21.5, and we're seeing a pretty major change in behavior. Under identical network load, 2.6.21 has a load average more than 3 times higher, cpu 0 spends well over 90% of its time in interrupts (vs ~30% under 2.6.17). When we hit 3k apache sessions, ksoftirqd eats 100% of cpu0 and our network traffic drops off rapidly. The end result is that 2.6.17 performs twice as well under this load. Can you oprofile it? # opreport -l CPU: AMD64 processors, speed 1994.52 MHz (estimated) Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 10 samples %app name symbol name 914379 48.8404 vmlinux-2.6.21.5 check_poison_obj 341920 18.2632 vmlinux-2.6.21.5 poison_obj I bet you have CONFIG_DEBUG_SLAB turned off in your 2.6.17 kernel, and turned on in your 2.6.21 kernel. -- Chris - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
Turn off slab debugging and preempt, and see if that helps. Jeff - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
> # opreport -l > CPU: AMD64 processors, speed 1994.52 MHz (estimated) > Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit > mask of 0x00 (No unit mask) count 10 > samples %app name symbol name > 914379 48.8404 vmlinux-2.6.21.5 check_poison_obj > 341920 18.2632 vmlinux-2.6.21.5 poison_obj You have slab debugging enabled. That makes everything slow. It costs you ~66% of your CPU time. Disable it. -Andi - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
Aaron Porter wrote: I'm in the process up upgrading a pool of apache servers from 2.6.17.8 to 2.6.21.5, and we're seeing a pretty major change in behavior. Under identical network load, 2.6.21 has a load average more than 3 times higher, cpu 0 spends well over 90% of its time in interrupts (vs ~30% under 2.6.17). When we hit 3k apache sessions, ksoftirqd eats 100% of cpu0 and our network traffic drops off rapidly. The end result is that 2.6.17 performs twice as well under this load. Is it always CPU 0, or does it move? Are you running irqbalance? If you're running irqbalance, you can run a script that alternates between 'cat /proc/interrupts' and 'mpstat -P ALL 5 10' and watch the offending interrupt jump around between processors. It's not as informative as oprofile, as Andi suggested, but it's really easy to set up. -- Chris - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
Aaron Porter <[EMAIL PROTECTED]> writes: > I'm in the process up upgrading a pool of apache servers from > 2.6.17.8 to 2.6.21.5, and we're seeing a pretty major change in behavior. > Under identical network load, 2.6.21 has a load average more than 3 times > higher, cpu 0 spends well over 90% of its time in interrupts (vs ~30% > under 2.6.17). When we hit 3k apache sessions, ksoftirqd eats 100% of cpu0 > and our network traffic drops off rapidly. The end result is that 2.6.17 > performs twice as well under this load. Can you oprofile it? -Andi - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
miserable performance of 2.6.21 under network load
I'm in the process up upgrading a pool of apache servers from 2.6.17.8 to 2.6.21.5, and we're seeing a pretty major change in behavior. Under identical network load, 2.6.21 has a load average more than 3 times higher, cpu 0 spends well over 90% of its time in interrupts (vs ~30% under 2.6.17). When we hit 3k apache sessions, ksoftirqd eats 100% of cpu0 and our network traffic drops off rapidly. The end result is that 2.6.17 performs twice as well under this load. - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
miserable performance of 2.6.21 under network load
I'm in the process up upgrading a pool of apache servers from 2.6.17.8 to 2.6.21.5, and we're seeing a pretty major change in behavior. Under identical network load, 2.6.21 has a load average more than 3 times higher, cpu 0 spends well over 90% of its time in interrupts (vs ~30% under 2.6.17). When we hit 3k apache sessions, ksoftirqd eats 100% of cpu0 and our network traffic drops off rapidly. The end result is that 2.6.17 performs twice as well under this load. - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
Aaron Porter [EMAIL PROTECTED] writes: I'm in the process up upgrading a pool of apache servers from 2.6.17.8 to 2.6.21.5, and we're seeing a pretty major change in behavior. Under identical network load, 2.6.21 has a load average more than 3 times higher, cpu 0 spends well over 90% of its time in interrupts (vs ~30% under 2.6.17). When we hit 3k apache sessions, ksoftirqd eats 100% of cpu0 and our network traffic drops off rapidly. The end result is that 2.6.17 performs twice as well under this load. Can you oprofile it? -Andi - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
Aaron Porter wrote: I'm in the process up upgrading a pool of apache servers from 2.6.17.8 to 2.6.21.5, and we're seeing a pretty major change in behavior. Under identical network load, 2.6.21 has a load average more than 3 times higher, cpu 0 spends well over 90% of its time in interrupts (vs ~30% under 2.6.17). When we hit 3k apache sessions, ksoftirqd eats 100% of cpu0 and our network traffic drops off rapidly. The end result is that 2.6.17 performs twice as well under this load. Is it always CPU 0, or does it move? Are you running irqbalance? If you're running irqbalance, you can run a script that alternates between 'cat /proc/interrupts' and 'mpstat -P ALL 5 10' and watch the offending interrupt jump around between processors. It's not as informative as oprofile, as Andi suggested, but it's really easy to set up. -- Chris - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
Turn off slab debugging and preempt, and see if that helps. Jeff - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
# opreport -l CPU: AMD64 processors, speed 1994.52 MHz (estimated) Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 10 samples %app name symbol name 914379 48.8404 vmlinux-2.6.21.5 check_poison_obj 341920 18.2632 vmlinux-2.6.21.5 poison_obj You have slab debugging enabled. That makes everything slow. It costs you ~66% of your CPU time. Disable it. -Andi - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
Aaron Porter wrote: On Tue, Jul 24, 2007 at 08:48:00PM +0200, Andi Kleen wrote: Aaron Porter [EMAIL PROTECTED] writes: I'm in the process up upgrading a pool of apache servers from 2.6.17.8 to 2.6.21.5, and we're seeing a pretty major change in behavior. Under identical network load, 2.6.21 has a load average more than 3 times higher, cpu 0 spends well over 90% of its time in interrupts (vs ~30% under 2.6.17). When we hit 3k apache sessions, ksoftirqd eats 100% of cpu0 and our network traffic drops off rapidly. The end result is that 2.6.17 performs twice as well under this load. Can you oprofile it? # opreport -l CPU: AMD64 processors, speed 1994.52 MHz (estimated) Counted CPU_CLK_UNHALTED events (Cycles outside of halt state) with a unit mask of 0x00 (No unit mask) count 10 samples %app name symbol name 914379 48.8404 vmlinux-2.6.21.5 check_poison_obj 341920 18.2632 vmlinux-2.6.21.5 poison_obj I bet you have CONFIG_DEBUG_SLAB turned off in your 2.6.17 kernel, and turned on in your 2.6.21 kernel. -- Chris - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
From: Aaron Porter [EMAIL PROTECTED] Date: Tue, 24 Jul 2007 11:49:09 -0700 samples %app name symbol name 914379 48.8404 vmlinux-2.6.21.5 check_poison_obj 341920 18.2632 vmlinux-2.6.21.5 poison_obj 37355 1.9953 nf_conntrack (no symbols) You have SLAB debugging enabled, turn it off if you want decent performance. - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: miserable performance of 2.6.21 under network load
On Tue, Jul 24, 2007 at 09:13:47PM +0200, Andi Kleen wrote: You have slab debugging enabled. That makes everything slow. It costs you ~66% of your CPU time. Disable it. Past peak for the day, but this does look like the culprit. Thanks to all who replied. - To unsubscribe from this list: send the line unsubscribe linux-kernel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/