From: "Tony 'Nicoya' Mantler" <nicoya@apia.dhs.org>
To: linux-kernel@vger.kernel.org
Subject: Disk performance irregularity in 2.4.x
Date: Wed, 25 Jun 2003 15:40:34 -0500 [thread overview]
Message-ID: <v04003a44bb1fb082955b@[192.168.1.4]> (raw)
Hello *,
I've been running 2.4.18 and 2.4.20 for a while on my server, and I've been
noticing a performance irregularity during read/write disk activity.
Usually the trigger is nothing more than reading and writing a constant
average of, say, 30-60K/s to disk. This will go normally most of the time,
but then, seemingly randomly, the machine will start writing to the disk at
a solid 0.5-1.0MB/s for 5 to 10 seconds while apparently not allowing any
other IO to take place (for example, my mp3s - playing on another machine
reading from the server via nfs - will stop playing for the duration of the
writeout). It can also be triggered more frequently by higher volume
read/writes (such as CVS activity).
I managed to catch one of these episodes in a vmstat trace just now during
some CVS activity. The first page is the base-level activity (1 second
intervals), then on the second page you can see the CVS activity starting
as the system starts reading from the disk. At about the bottom 1/3rd of
the second page the blackout starts as the kernel starts single-mindedly
writing data to disk, then on the 3rd page you can see where the normal CVS
activity resumes.
I am puzzled as to what is causing this, is it just bdflush tuning? Or is
something more bizzare going on? If it is bdflush tuning, what values do I
want to change?
Also, I have no idea where those 60000 extra context switches came from.
The main system drive is an IBM DeathStar 75GXP 60G wired up to the Promise
PCI card.
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
r b swpd free buff cache si so bi bo in cs us sy id wa
0 0 82056 4416 52108 163596 0 0 0 0 230 293 11 14 75 0
0 0 82056 4424 52108 163596 0 0 0 0 211 267 11 11 78 0
0 0 82056 4416 52108 163596 0 0 0 0 313 303 27 8 65 0
0 0 82056 4400 52124 163596 0 0 0 44 253 296 20 11 69 0
1 0 82056 4416 52108 163612 0 0 128 0 235 275 16 7 77 0
0 0 82056 4400 52108 163612 0 0 0 0 280 293 22 11 67 0
0 0 82056 4408 52108 163612 0 0 0 0 218 267 12 10 78 0
1 0 82056 4392 52124 163612 0 0 0 44 222 284 8 14 78 0
0 0 82056 4392 52124 163612 0 0 0 0 293 317 23 9 68 0
0 0 82056 4376 52104 163652 0 0 128 0 221 273 15 10 75 0
0 0 82056 4376 52104 163652 0 0 0 0 282 290 25 12 63 0
0 0 82056 4348 52132 163652 0 0 4 56 248 297 20 10 70 0
0 0 82056 4476 52132 163524 0 0 0 0 306 291 20 14 66 0
0 0 82056 4484 52132 163524 0 0 0 0 276 261 25 10 66 0
0 0 82056 4864 52080 163192 0 0 0 0 323 335 26 12 62 0
0 0 82056 4736 52080 163320 0 0 128 0 207 299 8 12 80 0
2 0 82056 4688 52116 163320 0 0 4 210 524 407 32 19 49 0
0 0 82056 4696 52116 163320 0 0 0 0 234 296 12 12 76 0
2 0 82056 4572 52116 163448 0 0 128 0 232 301 19 10 70 0
2 0 82056 4400 52080 163484 0 0 384 0 235 312 59 15 26 0
1 0 82056 4832 52028 163308 0 0 0 189 351 338 30 19 52 0
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
r b swpd free buff cache si so bi bo in cs us sy id wa
0 0 82056 4824 52028 163308 0 0 0 0 230 276 15 10 75 0
1 0 82056 4824 52028 163308 0 0 0 0 212 278 11 10 78 0
3 0 82056 4828 52028 163308 0 0 0 0 310 294 29 11 60 0
0 0 82056 4372 52068 163624 0 0 326 168 248 308 12 11 76 0
0 0 82056 4472 52116 163296 0 0 104 0 267 373 21 12 67 0
2 1 82056 4364 52156 163152 0 0 404 0 383 433 50 25 24 0
6 0 82056 4476 52344 162428 0 0 440 0 430 497 61 31 8 0
1 1 82056 4468 52432 162252 0 12 352 1047 1748 903 63 34 3 0
3 1 82056 4380 52448 162320 0 128 540 128 337 452 68 26 6 0
1 1 82056 4456 52752 162320 0 120 320 120 372 512 35 21 44 0
3 1 82056 4452 52920 162096 0 4 300 4 380 473 41 23 36 0
0 1 82056 4404 53204 161732 0 0 396 1556 381 453 29 25 45 0
3 0 82056 4356 53492 160876 0 0 328 0 379 482 25 35 40 0
2 1 82056 4412 53668 160352 0 0 292 0 419 524 33 25 42 0
1 1 82056 4420 53868 159744 0 0 232 0 346 413 29 33 38 0
1 3 82056 4408 54196 159256 0 0 312 2636 478 493 22 25 52 0
0 2 82056 4400 54148 159308 0 0 132 1644 451 331 24 11 64 0
1 1 82056 4428 54148 159308 0 0 128 964 414 316 11 15 74 0
3 1 82056 4424 54108 159352 0 0 260 716 373 348 27 12 61 0
1 2 82056 4400 54132 159352 0 0 0 1480 401 331 45 12 43 0
1 2 82056 4392 54068 159416 0 0 128 1452 427 311 12 13 75 0
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
r b swpd free buff cache si so bi bo in cs us sy id wa
0 1 82056 4388 54068 159416 0 0 0 1060 533 391 25 16 59 0
0 2 82056 4384 54068 159416 0 0 0 820 383 304 11 12 77 0
0 4 82056 4392 54056 159432 0 0 256 632 328 296 8 12 80 0
0 5 82056 4384 54060 159432 0 0 0 644 291 283 6 10 84 0
4 3 82056 4388 54064 159432 0 0 0 604 741 36406 23 20 57 0
5 1 82056 4472 54380 158804 0 0 700 32 1979 27335 58 42 0 0
1 1 82056 4468 54644 158396 0 0 348 324 638 683 46 26 28 0
1 1 82056 4408 54884 157908 0 0 372 0 412 559 26 24 50 0
1 1 82056 4436 55204 157204 0 0 384 0 360 495 21 27 52 0
2 1 82056 4432 55436 156728 0 0 524 0 435 553 32 30 39 0
0 1 82056 4460 55764 155792 0 0 380 884 396 533 21 27 52 0
ubb:~$ cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 5
model : 4
model name : Pentium MMX
stepping : 3
cpu MHz : 233.856
fdiv_bug : no
hlt_bug : no
f00f_bug : yes
coma_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr mce cx8 mmx
bogomips : 466.94
ubb:~$ lspci
00:00.0 Host bridge: ALi Corporation M1541 (rev 04)
00:01.0 PCI bridge: ALi Corporation M1541 PCI to AGP Controller (rev 04)
00:03.0 Bridge: ALi Corporation M7101 PMU
00:07.0 ISA bridge: ALi Corporation M1533 PCI to ISA Bridge [Aladdin IV]
(rev c3
)
00:09.0 SCSI storage controller: Adaptec AHA-2940/2940W / AIC-7871
00:0a.0 Ethernet controller: 3Com Corporation 3c905C-TX/TX-M [Tornado] (rev 6c)
00:0b.0 Unknown mass storage controller: Promise Technology, Inc. 20267
(rev 02)
00:0f.0 IDE interface: ALi Corporation M5229 IDE (rev c1)
01:00.0 VGA compatible controller: Trident Microsystems 3DImage 9750 (rev f3)
ubb:~$ uname -a
Linux ubb 2.4.20 #3 Sat May 17 22:57:31 CDT 2003 i586 GNU/Linux
ubb:~$ cat /proc/meminfo
total: used: free: shared: buffers: cached:
Mem: 329609216 324751360 4857856 0 52637696 178589696
Swap: 262168576 84054016 178114560
MemTotal: 321884 kB
MemFree: 4744 kB
MemShared: 0 kB
Buffers: 51404 kB
Cached: 159960 kB
SwapCached: 14444 kB
Active: 93736 kB
Inactive: 188572 kB
HighTotal: 0 kB
HighFree: 0 kB
LowTotal: 321884 kB
LowFree: 4744 kB
SwapTotal: 256024 kB
SwapFree: 173940 kB
ubb:~$ cat /proc/sys/vm/bdflush
30 500 0 0 500 3000 60 20 0
ubb:~$ mount
/dev/hda1 on / type ext3 (rw,errors=remount-ro)
proc on /proc type proc (rw)
devpts on /dev/pts type devpts (rw,gid=5,mode=620)
/dev/hda5 on /var type ext3 (rw,errors=remount-ro)
/dev/hda6 on /usr type ext3 (rw,errors=remount-ro)
/dev/hda7 on /home type ext3 (rw,errors=remount-ro)
ubb:~$ hdparm -Tt /dev/hda
/dev/hda:
Timing buffer-cache reads: 128 MB in 4.04 seconds = 31.68 MB/sec
Timing buffered disk reads: 64 MB in 5.47 seconds = 11.70 MB/sec
Cheers - Tony 'Nicoya' Mantler :)
--
Tony "Nicoya" Mantler - Renaissance Nerd Extraordinaire - nicoya@apia.dhs.org
Winnipeg, Manitoba, Canada -- http://nicoya.feline.pp.se/
reply other threads:[~2003-06-25 20:26 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='v04003a44bb1fb082955b@[192.168.1.4]' \
--to=nicoya@apia.dhs.org \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®