Date: Thu, 6 Sep 2007 20:48:53 +0200 From: Ulrich Spoerlein <uspoerlein@gmail.com> To: current@freebsd.org Cc: pjd@freebsd.org Subject: ZFS not caching right? Message-ID: <20070906184853.GB90021@roadrunner.spoerlein.net>
next in thread | raw e-mail | index | archive | help
Hi, it's me again with another stupid question,
ever since I switched my /usr and /home to ZFS, the system has become
_very_ unresponsive under load. This is not because of CPU load, but I/O
load.
This is on laptop with a lousy 4800RPM hard disk and 1 GB RAM. After
boot, the system is snappy as ever up to a certain point. If I build a
large port for example it takes less than an hour to bring the system to
a crawl.
Right now, I'm building OOo in the background and tried to start amarok.
This took, like 3 minutes or more while the disk is going nuts
# zpool iostat 1
tank 12.4G 14.8G 100 0 12.4M 0
tank 12.4G 14.8G 47 42 5.45M 387K
tank 12.4G 14.8G 27 85 3.34M 392K
tank 12.4G 14.8G 38 27 4.83M 129K
tank 12.4G 14.8G 47 46 5.94M 864K
tank 12.4G 14.8G 43 47 5.37M 242K
tank 12.4G 14.8G 84 0 10.2M 0
tank 12.4G 14.8G 63 0 7.92M 0
tank 12.4G 14.8G 87 0 10.9M 0
tank 12.4G 14.8G 73 0 9.16M 0
tank 12.4G 14.8G 35 72 4.46M 678K
tank 12.4G 14.8G 19 73 2.29M 997K
tank 12.4G 14.8G 71 3 8.91M 3.96K
tank 12.4G 14.8G 91 0 11.4M 0
tank 12.4G 14.8G 84 0 10.5M 0
tank 12.4G 14.8G 54 88 6.81M 403K
tank 12.4G 14.8G 77 0 9.65M 0
tank 12.4G 14.8G 82 0 10.3M 0
The mem and cache settings as shown by systat -vm are as follows
7 users Load 1.01 1.12 1.15 Sep 6 20:42
Mem:KB REAL VIRTUAL VN PAGER SWAP PAGER
Tot Share Tot Share Free in out in out
Act 515148 55612 1283636 118028 138308 count 6
All 713292 62660 3724916 139796 pages 2 27
Proc: Interrupts
r p d s w Csw Trp Sys Int Sof Flt 322 cow 370 total
4 1 143 3740 6830 20k 371 371 5797 4092 zfod 100 clk irq0
58 ozfod atkbd0 1
13.0%Sys 0.3%Intr 6.7%User 74.9%Nice 5.1%Idle 1%ozfod 128 rtc irq8
| | | | | | | | | | | daefr pcm0 iwi0+
=======>>>------------------------------------- 4016 prcfr 106 cbb0 bfe0+
3 dtbuf 4824 totfr 28 ata0 irq14
Namei Name-cache Dir-cache 50000 desvn 4 react 8 ata1 irq15
Calls hits % hits % 30904 numvn pdwak psm0 irq12
38614 38555 100 17491 frevn pdpgs
6 intrn
Disks ad0 da0 pass0 196884 wire
KB/t 8.36 0.00 0.00 552976 act
tps 27 0 0 131128 inact
MB/s 0.22 0.00 0.00 22796 cache
%busy 9 0 0 115512 free
110176 buf
Yesterday, the cache value was hovering at 900 and not improving much,
this is abnormal, right? I "tuned" ZFS with the following settings as
recommended by the Wiki and on -current
vfs.zfs.zil_disable="1"
vfs.zfs.prefetch_disable="1"
vfs.zfs.arc_max="128*1024*1024"
Is the tremendous amount of disk reading due to ZFS' nature or is it VM
related?
Cheers,
Ulrich Spoerlein
--
It is better to remain silent and be thought a fool,
than to speak, and remove all doubt.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070906184853.GB90021>