On Mar 25, 2012, at 6:51 PM, Aubrey Li wrote:
> On Mon, Mar 26, 2012 at 4:18 AM, Jim Mauro <james.ma...@oracle.com> wrote:
>> If you're chasing CPU utilization, specifically %sys (time in the kernel),
>> I would start with a time-based kernel profile.
>>
>> #dtrace -n 'profile-997hz /arg0/ { @[stack()] = count(); } tick-60sec {
>> trunc(@, 20); printa(@0; }'
>>
>> I would be curious to see where the CPU cycles are being consumed first,
>> before going down the lock path…
>>
>> This assume that most or all of CPU utilization is %sys. If it's %usr, we
>> take
>> a different approach.
>>
>
> Here is the output, I changed to "tick-5sec" and "trunc(@, 5)".
> No.2 and No.3 is what I care about.
>
> Thanks,
> -Aubrey
>
> 21 80536 :tick-5sec
> ========== 1 =====================
> genunix`avl_walk+0x6a
> genunix`as_gap_aligned+0x2b7
> unix`map_addr_proc+0x179
> unix`map_addr+0x8e
> genunix`choose_addr+0x9e
> zfs`zfs_map+0x161
> genunix`fop_map+0xc5
> genunix`smmap_common+0x268
> genunix`smmaplf32+0xa2
> genunix`syscall_ap+0x92
> unix`_sys_sysenter_post_swapgs+0x149
> 1427
>
> ========= 2 =================
> unix`mutex_delay_default+0x7
> unix`mutex_vector_enter+0x2ae
> zfs`zfs_zget+0x46
> zfs`zfs_root+0x55
> genunix`fsop_root+0x2d
> genunix`traverse+0x65
> genunix`lookuppnvp+0x446
> genunix`lookuppnatcred+0x119
> genunix`lookupnameatcred+0x97
> genunix`lookupnameat+0x6b
> genunix`vn_openat+0x147
> genunix`copen+0x493
> genunix`openat64+0x2d
> unix`_sys_sysenter_post_swapgs+0x149
> 2645
>
> ========== 3 =====================
> unix`mutex_delay_default+0x7
> unix`mutex_vector_enter+0x2ae
> zfs`zfs_zget+0x46
> zfs`zfs_root+0x55
> genunix`fsop_root+0x2d
> genunix`traverse+0x65
> genunix`lookuppnvp+0x446
> genunix`lookuppnatcred+0x119
> genunix`lookupnameatcred+0x97
> genunix`lookupnameat+0x6b
> genunix`cstatat_getvp+0x11e
> genunix`cstatat64_32+0x5d
> genunix`fstatat64_32+0x4c
> unix`_sys_sysenter_post_swapgs+0x149
> 3201
>
> ======== 4 ===========
> unix`i86_mwait+0xd
> unix`cpu_idle_mwait+0x154
> unix`idle+0x116
> unix`thread_start+0x8
> 3559
>
> ========= 5 ==============
> tmpfs`tmp_readdir+0x138
> genunix`fop_readdir+0xe8
> genunix`getdents64+0xd5
> unix`_sys_sysenter_post_swapgs+0x149
> 4589
>
> ========= 6 ================
> unix`strlen+0x3
> genunix`fop_readdir+0xe8
> genunix`getdents64+0xd5
> unix`_sys_sysenter_post_swapgs+0x149
> 5005
>
> =========== 7 =================
> tmpfs`tmp_readdir+0xc7
> genunix`fop_readdir+0xe8
> genunix`getdents64+0xd5
> unix`_sys_sysenter_post_swapgs+0x149
> 9548
>
>
> ========= 8 =======================
> unix`strlen+0x8
> genunix`fop_readdir+0xe8
> genunix`getdents64+0xd5
> unix`_sys_sysenter_post_swapgs+0x149
> 11166
>
>
> ============= 9 ===================
> unix`strlen+0xe
> genunix`fop_readdir+0xe8
> genunix`getdents64+0xd5
> unix`_sys_sysenter_post_swapgs+0x149
> 14491
>
> ============= 10 ==================
> tmpfs`tmp_readdir+0xbe
> genunix`fop_readdir+0xe8
> genunix`getdents64+0xd5
> unix`_sys_sysenter_post_swapgs+0x149
> 39540
Well, there you go. tmpfs is not zfs.
-- richard
--
ZFS and performance consulting
http://www.RichardElling.com
SCALE 10x, Los Angeles, Jan 20-22, 2012
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
