mmmmn

版权声明：转载时请以超链接形式标明文章原始出处和作者信息及本声明
http://mmmmn.blogbus.com/logs/2942168.html

1、Re: iostat -- is there something better?

-----------------------------------------------------------------------------

---

From: Robert Lawhead <news0000.5.unixguru@xxxxxxxxxxxxxxx>
Date: Tue, 13 Jun 2006 10:18:50 -0700

-----------------------------------------------------------------------------

---
Jay G. Scott wrote:

i've read iostat's man page, maybe that's what i did wrong.
if i do:
iostat disklist -l 20 whatever
it
1. doesn't give me everything in my disklist

Can you provide an example of the arguments you used and
output from iostat? Did the switches match the names?
(For example if your list contained c0t0d0s0, did you use -n).

2. still cuts the list off at 20

Isn't that why you used "-l 20"?

3, lists disks i don't care about.

depending on options, the disk goes on the left or the right....

And this matter why?
I think "-n" (names on right) is newer option, so perhaps
behavior is rooted in backwards compatibility.

w/o -l 20 i get far more output than i can digest. also, i can't get
things labelled by mount point.

Works for me, See below.
I think mountpoints are only given with "-n" output,
which seems sensible because names printed w/o -n don't
resemble anything from the output of "mount" or "df".

the thing just feels chaotic. is there something better?

For what specific purpose?

One thing that I think is a bug is iostat's interaction with /dev/fd.
For example I think if I have a file called 'disklist', that contains
the names of the devices to list, then

$ iostat -nxpm -l 10 <(cat disklist)

should work, and doesn't.

% iostat -nxpm # line numbers added by hand

extended device statistics
r/s w/s kr/s kw/s wait actv wsvc asvc %w %b device
0.8 0.3 18.9 5.2 0.0 0.0 9.7 4.5 0 0 c0t1d0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c0t1d0s0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c0t1d0s1
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c0t1d0s2
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c0t1d0s3
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c0t1d0s4
0.0 0.0 0.0 0.0 0.0 0.0 0.0 5.6 0 0 c0t1d0s5 (/export/stuff)
0.1 0.2 5.2 2.5 0.0 0.0 0.5 3.6 0 0 c0t1d0s6 (/export/home)
0.7 0.1 13.6 2.7 0.0 0.0 14.5 4.9 0 0 c0t1d0s7 (/usr/local)
0.2 0.1 8.5 1.3 0.0 0.0 0.0 7.4 0 0 c1t1d0s0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c1t1d0s7
1.8 0.9 23.3 21.8 0.1 0.0 42.5 11.4 1 2 c0t0d0
1.3 0.4 16.5 0.4 0.0 0.0 13.2 11.0 0 1 c0t0d0s0 (/)
0.0 0.0 0.0 0.0 0.0 0.0 22.5 24.9 0 0 c0t0d0s1
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c0t0d0s2
0.4 0.5 5.2 19.8 0.1 0.0 91.1 11.8 0 1 c0t0d0s3 (/var)
0.0 0.0 1.6 1.6 0.0 0.0 135.9 16.0 0 0 c0t0d0s4 (/opt)
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 fd0
0.2 0.1 8.6 1.3 0.0 0.0 0.0 6.3 0 0 c1t1d0
0.2 0.1 10.3 1.3 0.0 0.0 0.0 8.4 0 0 c1t2d0
0.2 0.1 10.3 1.3 0.0 0.0 0.0 9.7 0 0 c1t2d0s0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c1t2d0s7
0.0 0.0 0.0 0.0 0.0 0.0 0.0 5.2 0 0 c0t2d0
0.3 0.1 13.3 1.3 0.0 0.0 0.0 11.2 0 0 c1t3d0
0.3 0.1 13.2 1.3 0.0 0.0 0.0 12.7 0 0 c1t3d0s0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c1t3d0s7
0.3 0.1 13.4 1.3 0.0 0.0 0.0 11.1 0 0 c1t4d0
0.2 0.1 13.3 1.3 0.0 0.0 0.0 12.6 0 0 c1t4d0s0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 c1t4d0s7
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 st0

.

-----------------------------------------------------------------------------

---

References:
iostat -- is there something better?
From: Jay G. Scott
Prev by Date: Re: iostat -- is there something better?
Next by Date: Re: Acrobat reader
Previous by thread: Re: iostat -- is there something better?
Next by thread: Re: iostat -- is there something better?
Index(es):
Date
Thread

-----------------------------------------------------------------------------

---

Relevant Pages
Re: iostat -- is there something better?
... Jay G. Scott wrote: ... iostat disklist -l 20 whatever ... so
iostat added 20 additional disks ... (comp.unix.solaris)
Track process
... Below is the iostat information from one of the SUN servers ... (E4500

running
solaris 8 connecting to T3 via SAN switch). ... (comp.unix.solaris)
Re: Why does dd take so long on my Ultra 60
... Angelo Alvarez wrote: ... > starting to fail, so I want to replace
it as soon as possible. ... Does iostat -En show any errors against the drive

you're trying
to ... Scott ... (comp.sys.sun.hardware)

-----------------------------------------------------------------------------

---

We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(01)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06
-------
Re: iostat -- is there something better?

-----------------------------------------------------------------------------

---

From: "Daniel Rock" <v200624@xxxxxxxxxxx>
Date: Tue, 13 Jun 2006 18:08:57 +0000 (UTC)

-----------------------------------------------------------------------------

---
Jay G. Scott <gl@xxxxxxxxxxxxxxxxxxxxxxxx> wrote:

i've read iostat's man page, maybe that's what i did wrong.
if i do:
iostat disklist -l 20 whatever
it
1. doesn't give me everything in my disklist

Did you use the exact same names as in the "device" column?
"disklist" should also appear *after* any normal options:
iostat -l 20 disklist whatever

2. still cuts the list off at 20

Your disklist probably machted nothing, so iostat added 20 additional disks
to the output.

3, lists disks i don't care about.

See above.

If you use a disklist you should probably set -l 1 (or anything lower than
the number of disks in your disklist).

Example on my machine:

% iostat -xn
extended device statistics
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.1 0 0 c0d0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.3 0 0 c0d1
16.7 19.8 1143.1 75.7 0.8 0.1 21.1 3.7 5 8 c3d0
16.0 15.9 1137.6 71.8 0.8 0.2 25.3 4.7 5 10 c4d0
20.3 38.7 1219.0 152.2 0.9 0.2 15.4 3.1 5 12 c2d0
20.3 38.9 1218.6 152.3 0.9 0.2 15.8 3.1 5 12 c1d0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 fd0
6.8 1.1 44.3 5.2 0.0 0.1 0.4 10.6 0 3 md5
0.0 0.0 0.0 0.0 0.0 0.0 0.0 9.7 0 0 md12
0.0 0.0 0.0 0.0 0.0 0.0 0.0 10.1 0 0 md13
0.0 0.0 0.0 0.0 0.0 0.0 0.2 10.9 0 0 md14
3.4 1.1 22.2 5.2 0.0 0.1 0.0 11.4 0 2 md15
3.4 1.1 22.2 5.2 0.0 0.0 0.0 10.6 0 2 md19
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 server:vold(pid367)
% iostat -xn -l 1 c0d1 c1d0 fd0
extended device statistics
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.3 0 0 c0d1
20.3 38.9 1218.4 152.3 0.9 0.2 15.8 3.1 5 12 c1d0
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 fd0

--
Daniel
.

-----------------------------------------------------------------------------

---

Follow-Ups:
Re: iostat -- is there something better?
From: Jay G. Scott
References:
iostat -- is there something better?
From: Jay G. Scott
Prev by Date: Re: Acrobat reader
Next by Date: Re: Mount of rootfs(/) is being done in r/w mode
Previous by thread: Re: iostat -- is there something better?
Next by thread: Re: iostat -- is there something better?
Index(es):
Date
Thread

-----------------------------------------------------------------------------

---

Relevant Pages
Re: iostat -- is there something better?
... Jay G. Scott wrote: ... iostat disklist -l 20 whatever ... Did
the switches match the names? ... (comp.unix.solaris)
Track process
... Below is the iostat information from one of the SUN servers ... (E4500

running
solaris 8 connecting to T3 via SAN switch). ... (comp.unix.solaris)
SUMMARY: Identifying disks
... After working all night long my mind was not working straight so

obviously I ...
But I had tests fro last night so running iostat with option -n ... or just

take a look
and find disks you are looking for... ... so I do have paths and everything

else but I
need somehow connect them ... (SunManagers)
Re: iostat -- is there something better?
... Daniel Rock wrote: ... iostat disklist -l 20 whatever ... so iostat
added 20 additional disks ... grumble. ... (comp.unix.solaris)
Re: Disk labels (c0t0d0 & c1t0d => c2t0d0 & c3t0d0) weirdness
... Is there a Solaris 9 or 10 command that returns the same ... kind of
info as iostat -En but works for PC disks? ... (comp.unix.solaris)

-----------------------------------------------------------------------------

---

We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(06)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06      
----
Re: iostat -- is there something better?

-----------------------------------------------------------------------------

---

From: gl@xxxxxxxxxxxxxxxxxxxxxxxx (Jay G. Scott)
Date: Thu, 15 Jun 2006 16:04:06 +0000 (UTC)

-----------------------------------------------------------------------------

---
In article <e6muvp$18c6$1@xxxxxxxxxxxxxxx>,
Daniel Rock <v200624@xxxxxxxxxxx> wrote:

Jay G. Scott <gl@xxxxxxxxxxxxxxxxxxxxxxxx> wrote:

i've read iostat's man page, maybe that's what i did wrong.
if i do:
iostat disklist -l 20 whatever
it
1. doesn't give me everything in my disklist

Did you use the exact same names as in the "device" column?
"disklist" should also appear *after* any normal options:
iostat -l 20 disklist whatever

2. still cuts the list off at 20

Your disklist probably machted nothing, so iostat added 20 additional disks
to the output.

3, lists disks i don't care about.

See above.

If you use a disklist you should probably set -l 1 (or anything lower than
the number of disks in your disklist).

aaugh! that helps.

yours and the other posters comments helped, but there's still a problem:

lurch# iostat -mnxszM -l 1 `cat /tmp/rot3` 20 2
extended device statistics
r/s w/s Mr/s Mw/s wait actv wsvc_t asvc_t %w %b device
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.6 0 0 d19
0.0 0.3 0.0 0.0 0.0 0.0 7.7 11.5 0 0 d37 (/var)
0.0 0.6 0.0 0.0 0.0 0.0 9.6 28.2 1 1 d38 (/opt)
0.6 7.7 0.0 0.1 0.0 0.0 0.8 0.9 1 1 d44
0.0 1.2 0.0 0.0 0.0 0.0 4.8 0.7 1 0 d63

d19, for example, doesn't show the mount point. grumble.

but thanks to all.

(ahhh.... my complaining about where the disk label was, left, right,
is because if i were to write some script to digest the output it would
be harder to change options.)

j.

[snip]

--
Daniel

--
Jay Scott 512-835-3553 gl@xxxxxxxxxxxxxxxx
Head of Sun Support, Sr. Operating Systems Specialist
Applied Research Labs, Computer Science Div. S224
University of Texas at Austin
.

-----------------------------------------------------------------------------

---

References:
iostat -- is there something better?
From: Jay G. Scott
Re: iostat -- is there something better?
From: Daniel Rock
Prev by Date: Re: Solaris 8 system duplicate IP address
Next by Date: How to compile Apache 2.2.2 on Solaris 10 ?
Previous by thread: Re: iostat -- is there something better?
Next by thread: Memory for Ultra 45
Index(es):
Date
Thread

-----------------------------------------------------------------------------

---

Relevant Pages
SUMMARY: Identifying disks
... After working all night long my mind was not working straight so

obviously I ...
But I had tests fro last night so running iostat with option -n ... or just

take a look
and find disks you are looking for... ... so I do have paths and everything

else but I
need somehow connect them ... (SunManagers)
Re: Disk labels (c0t0d0 & c1t0d => c2t0d0 & c3t0d0) weirdness
... Is there a Solaris 9 or 10 command that returns the same ... kind of
info as iostat -En but works for PC disks? ... (comp.unix.solaris)
RE: [Q] how to check disk performance ?
... I'm sorry maybe I'm missing something here, ... >> We would like check
disks performance under AS2.1. ... > iostat is a pretty simple disk utility

which
is for some reason missing ... (RedHat)
RE: [Q] how to check disk performance ?
... > We would like check disks performance under AS2.1. ... > disks and SAN
disks. ... Sysstat - The sar, mpstat, and iostat commands ... (RedHat)

-----------------------------------------------------------------------------

---

We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(03)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06      
     

2、Re: Mount of rootfs(/) is being done in r/w mode

-----------------------------------------------------------------------------

---

From: Darren Dunham <ddunham@xxxxxxxxxxxxxxxx>
Date: Tue, 13 Jun 2006 16:52:01 GMT

-----------------------------------------------------------------------------

---
Imran <imran.1.618@xxxxxxxxx> wrote:

Hi all,

I was trying to write a service that checks attributes of rootfs before
it is
checked by fsck. The service has a dependency on boot-archive and has a
dependent "/system/filesystem/usr" service. It seems "/" is being
mounted in r/w mode before a
check is done on it. is that correct?

No, not normally.

Realize that the root filesystem initial (read-only) mount is
constructed directly by the kernel. There's no startup script that does
so, because it has to already be mounted to read the script.

If you're interested in '/', why do you have a dependency on usr?

When is the rootfs first mounted in the boot process?

A mount is constructed directly by the kernel as it boots.

It is mounted read/write by svc:/system/filesystem/root:default.

Is it ever
mounted in
read-only mode

yes. The initial kernel mount is read only. A pre-solaris 10 'boot -b'
will give you a shell before root is mounted read-write.

For Solaris 10, use 'boot -m milestone=none'.

I don't know what 'mount' is looking at. It will report / as
read/write, but in fact it is still read-only at that point.

or is Solaris fsck capable of checking filesystems that
are
mounted w/o damaging them?

no.

--
Darren Dunham ddunham@xxxxxxxx
Senior Technical Consultant TAOS http://www.taos.com/
Got some Dr Pepper? San Francisco, CA bay area
< This line left intentionally blank to confuse you. >
.

-----------------------------------------------------------------------------

---

Follow-Ups:
Re: Mount of rootfs(/) is being done in r/w mode
From: Andrew Gabriel
References:
Mount of rootfs(/) is being done in r/w mode
From: Imran
Prev by Date: Re: Memory for Ultra 45
Next by Date: Re: iostat -- is there something better?
Previous by thread: Mount of rootfs(/) is being done in r/w mode
Next by thread: Re: Mount of rootfs(/) is being done in r/w mode
Index(es):
Date
Thread

-----------------------------------------------------------------------------

---

Relevant Pages
Re: usb-storage && iRIVER flash player problem
... > The device is seen, but there is no way to mount it, nor re-partition

...
explain that the mountable vfat partition under linux is the ... > debugging

enabled
and not as modules, directly into the kernel. ... > nfs warning: ...

(Linux-Kernel)
Re: Getting CDRW to work in Debian
... Note that these instructions match a tuned modular kernel such as ...

method
to install your network and sound drivers. ... Test your configuration using

cdrecord
-scanbus. ... Create mount points for these devices. ... (Debian-User)
Re: XFS Mount Hangs the Partition (on latest kernel + many old 2.6.x ones)
... > attempting to mount a certain XFS partition (but not a different one on

the ...
> same hard-disk) caused the mount process to hang, ... > (that was using

kernel
2.4.15-rc2 from Linus), and then rebooted twice, ... (Linux-Kernel)
PROBLEM: (Bug report) USB Mass Storage in 2.6.0-test11 (cannot mount flash

drive)
... I cannot mount any USB mass storage drive when using kernel 2.6.0-test11

...
Dec 6 19:28:18 mycomp modprobe: FATAL: Module usb_storage already in kernel.

... Dec 6
14:49:25 mycomp kernel: Initializing USB Mass Storage driver... ...

4000-40ff: PCI CardBus
#02 ... (Linux-Kernel)
Re: evangelizing linux not needed
... Both should be umask (do a "man mount" to see mounting options). ...
an installer to guess things for you. ... used for your stock kernel. ...
(alt.os.linux)

-----------------------------------------------------------------------------

---

We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(08)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06      
----
Re: Mount of rootfs(/) is being done in r/w mode

-----------------------------------------------------------------------------

---

From: Darren Dunham <ddunham@xxxxxxxxxxxxxxxx>
Date: Wed, 14 Jun 2006 23:50:34 GMT

-----------------------------------------------------------------------------

---
Andrew Gabriel <andrew@xxxxxxxxxxxxxxxxxxxx> wrote:

For Solaris 10, use 'boot -m milestone=none'.

I don't know what 'mount' is looking at. It will report / as
read/write, but in fact it is still read-only at that point.

Maybe you have an old /etc/mnttab file underneath the mnttab
mount point, and mnttab wasn't yet mounted?

I thought of that right after I got the machine back up, but didn't want
to set everything back up again. So I finally got around to trying
another test and that doesn't seem to be the case...

# umount /etc/mnttab
# df -k
# ls -l /etc/mnttab
-r--r--r-- 1 root root 0 Jan 21 2005 /etc/mnttab

So it really is the kernel view of the mounts, not a stale file.

The "early" mnttab has no flag indications at all (neither rw or ro).

# head -1 /etc/mnttab
/pci@1f,0/pci@1,1/ide@3/disk@0,0:a / ufs dev=2200008 0

So I think 'mount' just assumes that no flags equals read/write (which
isn't true in this case).

In fact.... I don't know how to mount this filesystem (explicitly) read
only. Now that I look, the mount_ufs man page doesn't show a 'ro'
option, and my attempts to do so fail.

# mount -r -F ufs -o remount /dev/dsk/c0t0d0s0 /
mount: /dev/dsk/c0t0d0s0 is not this fstype
# mount -F ufs -o remount,ro /dev/dsk/c0t0d0s0 /
mount: /dev/dsk/c0t0d0s0 is not this fstype
but...
# mount -F ufs -o remount /dev/dsk/c0t0d0s0 /
#

"normal" filesystems can be mounted read-only, so why were the ro | rw
options removed from the mount_ufs man page?

--
Darren Dunham ddunham@xxxxxxxx
Senior Technical Consultant TAOS http://www.taos.com/
Got some Dr Pepper? San Francisco, CA bay area
< This line left intentionally blank to confuse you. >
.

-----------------------------------------------------------------------------

---

References:
Mount of rootfs(/) is being done in r/w mode
From: Imran
Re: Mount of rootfs(/) is being done in r/w mode
From: Darren Dunham
Re: Mount of rootfs(/) is being done in r/w mode
From: Andrew Gabriel
Prev by Date: T2000 running Clearcase
Next by Date: Re: Acrobat reader
Previous by thread: Re: Mount of rootfs(/) is being done in r/w mode
Next by thread: Customization Of Java Desktop
Index(es):
Date
Thread

-----------------------------------------------------------------------------

---

Relevant Pages
Re: HFS Plus R/W Mounting Problem
... > I'm trying to mount my iPod on FC2 as a read/write hfs plus file ... My
fstab entry looks like: ... > Sometimes it mounts as read/write but other

times
its just read only. ... What kernel are you using? ... (Fedora)
Re: mount & umount read-only floppy: unmount failed: Input/output error ??
... Rob wrote: ... if the mount didn't work. ... CAN read/write
mount a read/only-floppy. ... Unless, you take the floppy out, change it ...
(freebsd-questions)
HFS Plus R/W Mounting Problem
... I'm trying to mount my iPod on FC2 as a read/write hfs plus file system.

...
fstab entry looks like: ... (Fedora)
RE: Freebsd ufs under WIN
... > for read/write. ... > I am almost 100% certain that the real question
was: ... >> not only can I read and write to them, I can mount them as ...
is that I can *map* Samba shares to Network drives, ... (freebsd-questions)
Re: why vnode as readonly fs?
... > sam wrote: ... > You can't mount a cd9660 fs read/write.
... > matter of mdconfig but of the fs itself. ...

(comp.unix.bsd.freebsd.misc)

-----------------------------------------------------------------------------

---

We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(08)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06      

3、Re: 3310 RAID1 (mirror) Question

-----------------------------------------------------------------------------

---

From: ptribble@xxxxxxxxxxx (Peter Tribble)
Date: 13 Jun 2006 08:49:50 GMT

-----------------------------------------------------------------------------

---
In article <1150166605.509903.285610@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
"BC" <unix_ed@xxxxxxxxx> writes:

Question about setting up a Storedge 3310 - mirror. The 3310 has 12
drives, 2 channels. When creating the logical drive, I select 5 drives
from each channel (10 drives total) and select RAID1 (2 drives will be
saved for Global spares). I understand that the array will configure
RAID 1+0 and is also smart enough to setup the mirrored pairs across
both channels as described in this short Sun Doc:
http://sunsolve.sun.com/search/document.do?assetkey=1-9-71182-1

3 hrs later, after the drives are initialized, I see one logical drive,
but instructions are not clear how to setup the Solaris 9 OS to use
both controllers? I'm not sure how the dual channels are presented to
the OS? Or does the array deal with it all?

The 3310 will only give you a single host connection. (You could
connect another host to the array, or even the same host more than
once, but there isn't any way to get path redundancy and failover.)

Given two controllers, if one fails then it will just use the
other. Automatically and transparently. Works great. No interruption to
service.

(The downside to this is that you have to actively monitor the array to
check that both controllers are running, because the system might not
notice at all if one fails.)

Should I have setup 5 drives (channel 1) as RAID 0 and the other 5
drives as RAID 0 (channel 3), presented them both to the server and
used SVM to mirror across the controllers?

Definitely not. If you lose a disk in a stripe you lose the whole
stripe and all your data redundancy goes. If you want 0+1 to turn into
1+0, you need to do it all inside the array or all inside SVM.

Server is a V480 (Solaris 9), 3310 array firmware version 3.25, Fiber
channel

That doesn't make sense. The 3310 is a SCSI array, if you had Fiber
Channel it would be a 3510 (given the channels IDs you mention, it
sounds like you do have a SCSI based 3310). But if you did have a 3510
then my notes are here:

http://www.petertribble.co.uk/Solaris/3510.html
http://www.petertribble.co.uk/Solaris/mpxio.html

--
-Peter Tribble
L.I.S., University of Hertfordshire - http://www.herts.ac.uk/
http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/
.

-----------------------------------------------------------------------------

---

Follow-Ups:
Re: 3310 RAID1 (mirror) Question
From: BC
References:
3310 RAID1 (mirror) Question
From: BC
Prev by Date: Re: Problem with SSH
Next by Date: Re: Acrobat reader
Previous by thread: 3310 RAID1 (mirror) Question
Next by thread: Re: 3310 RAID1 (mirror) Question
Index(es):
Date
Thread

-----------------------------------------------------------------------------

---

Relevant Pages
Re: 3310 RAID1 (mirror) Question
... connect another host to the array, or even the same host more than ... I

select
5 drives ... I understand that the array will configure ... both channels
as described in this short Sun Doc: ... (comp.unix.solaris)
Re: Single P4, many IDE PCI cards == trouble??
... * Promise 20268 PCI cards ... Only 3 drives, ... and one disk
to each of the channels of the SI680 ... (Linux-Kernel)
Re: New Hd Install - No Udma Support, Dma 2 Only
... and secondary channels in device manager as well as the ide controller.

...
No listing of UDMA type. ... Does your motherboard really support large hard

drives?
... (alt.comp.hardware.pc-homebuilt)
Re: Booting
... my problem is that I have three hard drives all set ... All on their own

channels
and so forth set to master. ... > the problem because they are all master

drives, ...
(microsoft.public.windowsxp.general)
3310 RAID1 (mirror) Question
... I select 5 drives ... RAID 1+0 and is also smart enough to setup the

mirrored
pairs across ... both channels as described in this short Sun Doc: ... used
SVM to mirror across the controllers? ... (comp.unix.solaris)

-----------------------------------------------------------------------------

---

We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(10)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06      
----
Re: 3310 RAID1 (mirror) Question

-----------------------------------------------------------------------------

---

From: ptribble@xxxxxxxxxxx (Peter Tribble)
Date: 13 Jun 2006 13:58:37 GMT

-----------------------------------------------------------------------------

---
In article <1150204064.354224.221870@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
"BC" <unix_ed@xxxxxxxxx> writes:

Thanks for the help Peter,
You're right, the 3310 is SCSI and not fiber. (You know how it goes
whenit's late).

Yup. ;-)

you stated:

The 3310 will only give you a single host connection. (You could
connect another host to the array, or even the same host more than
once, but there isn't any way to get path redundancy and failover.)

Here's what I see with a format though:

0. c1t0d0 <SEAGATE-ST373307FSUN72G-0307 cyl 14087 alt 2 hd 24
sec 424>
/pci@9,600000/SUNW,qlc@2/fp@0,0/ssd@w2100000c50ebb5f7,0
1. c1t1d0 <SUN72G cyl 14087 alt 2 hd 24 sec 424>
/pci@9,600000/SUNW,qlc@2/fp@0,0/ssd@w21000011c62a46e2,0
2. c4t1d0 <SUN-StorEdge3310-0325 cyl 65533 alt 2 hd 64 sec 341>
/pci@8,700000/pci@3/scsi@5/sd@1,0
3. c6t1d0 <SUN-StorEdge3310-0325 cyl 65533 alt 2 hd 64 sec 341>
/pci@8,700000/pci@2/scsi@5/sd@1,0

So you have two SCSI connections to the array?

4. c8t40d0 <SUN StorEdge 3511 327R 91647800>
/pci@8,700000/SUNW,qlc@4/fp@0,0/ssd@w216000c0ff80670d,0

Items 2 and 3. I have two LUN's - 682GB each. I assume one LUN from
each side of the 3310. Based on what you stated, I would expect to
see one LUN. Do I now use SVM and create submirrors between 2 and 3?

How many logical drives have you created? You say you've created just
the one (mirror with 10 drives - which would be the right size if you
have 146G drives). What does the sccli "show ld" command give you?

If you've just created the one logical drive then you're just seeing
the same LUN twice - by different paths. In that case, really bad
things can happen if you try and access it from both paths. (As far as
I know, it isn't possible to use mpxio to use both paths in the same
way you can for a fiber array.)

--
-Peter Tribble
L.I.S., University of Hertfordshire - http://www.herts.ac.uk/
http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/
.

-----------------------------------------------------------------------------

---

Follow-Ups:
Re: 3310 RAID1 (mirror) Question
From: BC
References:
3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: Peter Tribble
Re: 3310 RAID1 (mirror) Question
From: BC
Prev by Date: Re: Sun Google
Next by Date: Re: Sun Google
Previous by thread: Re: 3310 RAID1 (mirror) Question
Next by thread: Re: 3310 RAID1 (mirror) Question
Index(es):
Date
Thread

-----------------------------------------------------------------------------

---

Relevant Pages
Re: IBM FastT vs. EMC Clarion
... AIX hosts have to be shutdown when doing the firmware upgrade ... any AIX
servers attached to that unit will have to be shutdown. ... interrupting host

I/O access to
the array. ... interrupting host I/O access to the array thousands of times

with
many ... (AIX-L)
Re: 3310 RAID1 (mirror) Question
... connect another host to the array, or even the same host more than ... I

select
5 drives ... I understand that the array will configure ... both channels
as described in this short Sun Doc: ... (comp.unix.solaris)
Re: Problem with filesystem on RAID5 array (fsck -b?)
... Dell PowerVault 220S disk cabinet. ... Two days ago a disk failed in
one of the arrays. ... Rebuilding the array ... delete logical drives

supported
by controller ... (freebsd-stable)
Re: Newbie: graphics cards
... I am using an IBM 3H/3L SCSI Raid Controller Card and the system sees

that, but does
not see my RAID Array. ... have you booted the ServeRAID support CD to set up

...
your array and logical drives? ... (comp.sys.ibm.pc.hardware.misc)
Re: Mirror two A1000 with two (non-cluster) nodes
... Each array has two RAID-5 volumes, for a total of 4 volumes: ...

Controller
A owns host 1's ... while Controller B owns host 2's volume. ... You can
use pigtail adapters, where the connector is on one side of the ...

(comp.sys.sun.admin)

-----------------------------------------------------------------------------

---

We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(02)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06      
-------
Re: 3310 RAID1 (mirror) Question

-----------------------------------------------------------------------------

---

From: ptribble@xxxxxxxxxxx (Peter Tribble)
Date: 13 Jun 2006 22:03:18 GMT

-----------------------------------------------------------------------------

---
In article <1150228835.650805.248580@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
"BC" <unix_ed@xxxxxxxxx> writes:

RE: sccli
I downloaded and installed sccli, but can't get info on the 3310
because it's not the primary controller. Most of the sccli commands
require the primary controller. As I showed in previous output a few
posts back, the 3310 configured itself on the secondary controller. In
the "Edit Logical Drives" section, there is a submenu named "Logical
Drive Assignments". It has an option to:
"Redundant Controller Logical Drive Assign to Primary Controller?" Yes
or No

when I select Yes, it changes my host Luns so they are on the primary
and not the secondary chl- but requires a controller reset. I was a
bit afraid it was going to change my disk logical device names in the
OS, so I bailed out of it and set things back to the secondary
controller.

My question is: If I change to Primary (so I can use sccli sub
commands), will my logical device name for the drive (as it appears in
the format utility) change? Thus, destroying my exisiting file systems
and soft partitions?

Now, I don't know what's going to happen. I've never tried what you're
thinking of doing, so I'm not going to do anything more than take a
guess based on my understanding of how the thing works.

My guess is that if you move it from the secondary to the primary
controller then it will change the ID from t1 to t0 (so c4t1d0 becomes
c4t0d0). It won't destroy any filesystems or anything like that, but
you would have to go through the system so that Solaris knows to look
in a different place.

--
-Peter Tribble
L.I.S., University of Hertfordshire - http://www.herts.ac.uk/
http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/
.

-----------------------------------------------------------------------------

---

Follow-Ups:
Re: 3310 RAID1 (mirror) Question
From: Joe D.
References:
3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: Peter Tribble
Re: 3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: Peter Tribble
Re: 3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: Peter Tribble
Re: 3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: BC
Prev by Date: Re: Customization Of Java Desktop
Next by Date: Re: changing routing table preference
Previous by thread: Re: 3310 RAID1 (mirror) Question
Next by thread: Re: 3310 RAID1 (mirror) Question
Index(es):
Date
Thread

-----------------------------------------------------------------------------

---

Relevant Pages
Re: 3310 RAID1 (mirror) Question
... I downloaded and installed sccli, but can't get info on the 3310 ...

because
it's not the primary controller. ... On my 3510s with two logical drives, ...
(comp.unix.solaris)
MVP? how do I restore video during boot
... the system wanted to treat the primary controller as the secondary

controller
... So when we boot - all goes well from the bios boot through windows boot

...
(microsoft.public.windowsxp.perform_maintain)
Re: 3310 RAID1 (mirror) Question
... Just install the sccli software on the attached host. ... is on the

primary
controller. ... On my 3510s with two logical drives, ... assigned one to each
controller. ... (comp.unix.solaris)
Configure multipathing in 3510 with V440
... First logical disk I assigned to primary controller. ... But when I

remove
primary FCAL channel cable from primary controller. ... (SunManagers)
Re: Sandisk fixed disk utility error #20
... (for a disk on the primary controller) ... atcfwchg /P /F ...
(for a disk on the secondary controller) ...

(microsoft.public.windowsxp.embedded)

-----------------------------------------------------------------------------

---

We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(02)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06      
------
Re: 3310 RAID1 (mirror) Question

-----------------------------------------------------------------------------

---

From: ptribble@xxxxxxxxxxx (Peter Tribble)
Date: 13 Jun 2006 22:03:18 GMT

-----------------------------------------------------------------------------

---
In article <1150228835.650805.248580@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
"BC" <unix_ed@xxxxxxxxx> writes:

RE: sccli
I downloaded and installed sccli, but can't get info on the 3310
because it's not the primary controller. Most of the sccli commands
require the primary controller. As I showed in previous output a few
posts back, the 3310 configured itself on the secondary controller. In
the "Edit Logical Drives" section, there is a submenu named "Logical
Drive Assignments". It has an option to:
"Redundant Controller Logical Drive Assign to Primary Controller?" Yes
or No

when I select Yes, it changes my host Luns so they are on the primary
and not the secondary chl- but requires a controller reset. I was a
bit afraid it was going to change my disk logical device names in the
OS, so I bailed out of it and set things back to the secondary
controller.

My question is: If I change to Primary (so I can use sccli sub
commands), will my logical device name for the drive (as it appears in
the format utility) change? Thus, destroying my exisiting file systems
and soft partitions?

Now, I don't know what's going to happen. I've never tried what you're
thinking of doing, so I'm not going to do anything more than take a
guess based on my understanding of how the thing works.

My guess is that if you move it from the secondary to the primary
controller then it will change the ID from t1 to t0 (so c4t1d0 becomes
c4t0d0). It won't destroy any filesystems or anything like that, but
you would have to go through the system so that Solaris knows to look
in a different place.

--
-Peter Tribble
L.I.S., University of Hertfordshire - http://www.herts.ac.uk/
http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/
.

-----------------------------------------------------------------------------

---

Follow-Ups:
Re: 3310 RAID1 (mirror) Question
From: Joe D.
References:
3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: Peter Tribble
Re: 3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: Peter Tribble
Re: 3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: Peter Tribble
Re: 3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: BC
Prev by Date: Re: Customization Of Java Desktop
Next by Date: Re: changing routing table preference
Previous by thread: Re: 3310 RAID1 (mirror) Question
Next by thread: Re: 3310 RAID1 (mirror) Question
Index(es):
Date
Thread

-----------------------------------------------------------------------------

---

Relevant Pages
Re: 3310 RAID1 (mirror) Question
... I downloaded and installed sccli, but can't get info on the 3310 ...

because
it's not the primary controller. ... On my 3510s with two logical drives, ...
(comp.unix.solaris)
MVP? how do I restore video during boot
... the system wanted to treat the primary controller as the secondary

controller
... So when we boot - all goes well from the bios boot through windows boot

...
(microsoft.public.windowsxp.perform_maintain)
Re: 3310 RAID1 (mirror) Question
... Just install the sccli software on the attached host. ... is on the

primary
controller. ... On my 3510s with two logical drives, ... assigned one to each
controller. ... (comp.unix.solaris)
Configure multipathing in 3510 with V440
... First logical disk I assigned to primary controller. ... But when I

remove
primary FCAL channel cable from primary controller. ... (SunManagers)
Re: Sandisk fixed disk utility error #20
... (for a disk on the primary controller) ... atcfwchg /P /F ...
(for a disk on the secondary controller) ...

(microsoft.public.windowsxp.embedded)

-----------------------------------------------------------------------------

---

We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(02)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06      
----
Re: 3310 RAID1 (mirror) Question

-----------------------------------------------------------------------------

---

From: ptribble@xxxxxxxxxxx (Peter Tribble)
Date: 14 Jun 2006 16:13:59 GMT

-----------------------------------------------------------------------------

---
In article <1150288691.925660.262540@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
"Joe D." <newbie_from_newbie@xxxxxxxxx> writes:

Good morning, gentlemen! A very interesting post.

I am about to set up a 3310 (probably a 3320, actually, but I presume
they are essentially the same as far as config goes?).

Pretty well. The 3320 is Ultra320, while the 3310 is Ultra160. And they
do take different firmware (new patches for both were released today)
so don't try applying 3310 firmware to a 3320 (or vice versa). But as
far as configuration goes I haven't noticed any differences.

I presume from
the post that you both have the RAID controller option. We're not going
to go with the RAID controller, due to cost considerations. Rather, I
was planning on simply using SVM for the system's boot disks, and use
the 3310's disks via raw access for a Sybase database.

I wonder if I could trouble you to elaborate a little on some points
made earlier:

Peter Said:
The 3310 will only give you a single host connection. (You could
connect another host to the array, or even the same host more than
once, but there isn't any way to get path redundancy and failover.)

Given two controllers, if one fails then it will just use the
other. Automatically and transparently. Works great. No interruption to
service.

and later when BD noted that he had 2 SCSI connections to the array and
was seeing 2 disks:

.... disconnect the SCSI cable from the
host. You don't really want this to be accessible, as it's just asking
for trouble.

One of the things I was hoping for in purchasing these arrays was the
dual-path redundancy (which it sounds like I'm not gonna get).

Well, SCSI is intrinsically single-path - unlike FC disks which are
dual-ported. So with JBOD you could have one path to a disk and a
different path to its mirrored (on the host) partner.

Is this
to say then, that even if I order a pair of PCI SCSI cards that I
shouldn't connect both of them?

With a RAID controller, you shouldn't. With JBOD, you should, and split
the busses so you attach half the disks to one card and half to the other.

My past experience had been with
Veritas which recognized automatically 2 paths to the same disk(s) and
presented you with only the one, even though you could see both from
format.

That's a good point. I hadn't thought about Veritas (previous
experience makes my mind blank it out - oh the pain!). I don't actually
know what Veritas would do with it.

I realize we may be talking apples and oranges here since you're using
the RAID controller (again, I'm presuming) and we won't be, and also
we'll be accessing them as raw devices, but is what I'm planning
feasible/recommended? Any alarm bells going off in your heads?

If you take out the raid controller, then I would run with a dual-bus
configuration. Connect one half of the disks to one card, the other
half to the other card, and then create mirrors (either SVM or within
Sybase) between disks on the two controllers.

In other words, split it in half and treat it as two 6-disk arrays that
happen to be in one box.

Note that the rules for JBOD and the rules for RAID are different.

--
-Peter Tribble
L.I.S., University of Hertfordshire - http://www.herts.ac.uk/
http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/
.

-----------------------------------------------------------------------------

---

Follow-Ups:
Re: 3310 RAID1 (mirror) Question
From: Joe D.
References:
3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: Peter Tribble
Re: 3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: Peter Tribble
Re: 3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: Peter Tribble
Re: 3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: BC
Re: 3310 RAID1 (mirror) Question
From: Peter Tribble
Re: 3310 RAID1 (mirror) Question
From: Joe D.
Prev by Date: Re: Faster backups - possibly Solaris pipe buffer size
Next by Date: Power Management Problems (DPMS)
Previous by thread: Re: 3310 RAID1 (mirror) Question
Next by thread: Re: 3310 RAID1 (mirror) Question
Index(es):
Date
Thread

-----------------------------------------------------------------------------

---

Relevant Pages
MPXIO: Cant get to encapsulate drives ...
... I can't get MPXIO to encapsulate the following pairs of disks ... are

derived
from a Clariion FC4700 unit (through two Brocade ... # Load balancing global

configuration:
... # List of ULP modules for loading during port driver attach time ...

(comp.unix.solaris)
Re: Flashing lights on my disk drives on an Alphaserver ES40
... When you have internal disks a junk card is installed in the ES40 -

assuming ...
> or whatever the controller is - even if it is not a RAID controller it>

should be listed
in SRM e.g. a PKxn device. ... > If the RAID controller is seen by SRM, go

into the RAID config
utility> and check that your physical drives are seen and your> logical

drives are still
defined ... > AND a RAID card, it might be normal to have no devices showing

on the>
internal SCSI bus, so if the RAID card disappears then there would> be

'unused' controllers - make
sure you follow the disk cables back! ... (comp.os.vms)
Re: Flashing lights on my disk drives on an Alphaserver ES40
... When you have internal disks a junk card is installed in the ES40 -

assuming ...
> or whatever the controller is - even if it is not a RAID controller it>

should be listed
in SRM e.g. a PKxn device. ... > If the RAID controller is seen by SRM, go

into the RAID config
utility> and check that your physical drives are seen and your> logical

drives are still
defined ... > AND a RAID card, it might be normal to have no devices showing

on the>
internal SCSI bus, so if the RAID card disappears then there would> be

'unused' controllers - make
sure you follow the disk cables back! ... (comp.sys.dec)
Re: Odd news transit performance problem
... > box is idle - there's no significant contention for memory, disk, ...

The
host is fed news from only two sources which on average equates to 25 ...

This still only
pushes the disks to 50% busy. ... 11424 Kbytes allocated to network ...

(freebsd-performance)
Re: ext3 filesystem not working with raid5
... and the array might already have reconstructed with bad data due ...

consider
*maxtor*250GIG IDE disks to be nice for single-user ... -rf' or a rooted

system or a RAID
controller running amuck at startup. ... > exactly what the problem is with

the ext3
filesystem and how to go ... (alt.linux)

-----------------------------------------------------------------------------

---

We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(02)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06      

4、Re: Optimizing memory usage of high performance webserver

-----------------------------------------------------------------------------

---

From: Rich Teer <rich.teer@xxxxxxxxxxxxxx>
Date: Sun, 11 Jun 2006 18:28:36 GMT

-----------------------------------------------------------------------------

---
On Sun, 11 Jun 2006, fredje110@xxxxxxxxxxx wrote:

We are running 4 Sun Fire V440 with 4 CPU's and 8 GB RAM as Apache
webservers. We do not use a dedicated caching/SQUID tier and we do not
have dedicated image servers (the amount of pictures on our website is
very large). It should not be any surprise that our internal discs have
reached their I/O limits.

iostat will give you some more info here.

On the other site the output of top shows the following:

Memory: 8192M real, 4892M free, 177M swap in use, 9538M swap free

It would make a big difference if we could load 3-4 GB of cached
content into memory (RamDisk?)

Top isn't the best tool to use on Solaris (use prstat instead), but
more to the point, if your files are static, they'll be cached in
RAM by the kernel anyway (more pressing needs for RAM nothwithstanding).

Our web tier is running Sun Solaris 8. How could we tune/convince
Solaris to load more cached content into memory and reduce disk I/O.

Use performance analysis tools like iostat to make sure you are actually
disk I/O bound before you do any tuning. But I will say this: you will
very likely see a significant performance increase if you upgrade those
servers to Solaris 10. Upgrading to Solaris 10 will also enable you to
use DTrace to help identify any performance bottlenecks. (If you can
wait a week or two, you'll be able to use SOlaris 10 6/06, which comes
with ZFS...)

HTH,

--
Rich Teer, SCNA, SCSA, OpenSolaris CAB member

President,
Rite Online Inc.

Voice: +1 (250) 979-1638
URL: http://www.rite-group.com/rich
.

-----------------------------------------------------------------------------

---

Follow-Ups:
Re: Optimizing memory usage of high performance webserver
From: KJ
Re: Optimizing memory usage of high performance webserver
From: Chris Lawson
Re: Optimizing memory usage of high performance webserver
From: Daniel Rock
References:
Optimizing memory usage of high performance webserver
From: fredje110
Prev by Date: Re: New server to play on
Next by Date: Re: Optimizing memory usage of high performance webserver
Previous by thread: Optimizing memory usage of high performance webserver
Next by thread: Re: Optimizing memory usage of high performance webserver
Index(es):
Date
Thread
We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(09)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06      
----
Re: Optimizing memory usage of high performance webserver

-----------------------------------------------------------------------------

---

From: "victor.engle@xxxxxxxxx" <victor.engle@xxxxxxxxx>
Date: 11 Jun 2006 13:30:59 -0700

-----------------------------------------------------------------------------

---
Daniel,

I agree with Rich's comments. We need to see the iostat data for those
drives housing your content. Also some network and cpu statistics and
maybe some vmstat data captured during a time when the systems is
experiencing the performance issue.

The V440 I believe has the Sparc IIIi processors so depending on how
much network bandwidth you're trying to support the system could even
be CPU bound at times. Is the content storage direct attached? What
kind of storage is it. A fibre channel SAN of course would be very
expensive but if you have to have faster storage then a midrange SAN
solution might be a reasonable thing to consider but I wouldn't insist
on active/active. That would likely add significant cost and it is
really hard to believe it's needed for a web server. You can configure
an active passive array such that you can utilize 2 or more HBA's on
the host side. I can provide details if needed.

Also, a selling point for the SAN idea is that it would open up shared
storage clustering possibilities for you.

Good luck with resolving the problem.

Regards,

Vic Engle

Daniel Rock wrote:

Rich Teer <rich.teer@xxxxxxxxxxxxxx> wrote:

On Sun, 11 Jun 2006, fredje110@xxxxxxxxxxx wrote:

Memory: 8192M real, 4892M free, 177M swap in use, 9538M swap free

Top isn't the best tool to use on Solaris (use prstat instead),

prstat won't display anything similar to that above (physical memory usage),
so why use it if it cannot give me the information I'm interested in?

--
Daniel

.

-----------------------------------------------------------------------------

---

References:
Optimizing memory usage of high performance webserver
From: fredje110
Re: Optimizing memory usage of high performance webserver
From: Rich Teer
Re: Optimizing memory usage of high performance webserver
From: Daniel Rock
Prev by Date: Re: Optimizing memory usage of high performance webserver
Next by Date: Re: Optimizing memory usage of high performance webserver
Previous by thread: Re: Optimizing memory usage of high performance webserver
Next by thread: Re: Optimizing memory usage of high performance webserver
Index(es):
Date
Thread

-----------------------------------------------------------------------------

---

Relevant Pages
Re: Information store on WSS 2003
... communication with the storage, not with the network. ... Check for

firmware
and driver updates for storage solution. ... Check with your storage

manufacturer to see
if one or more settings ... Please do not send email directly to this alias.

...
(microsoft.public.exchange.admin)
Re: [SLE] Why logical volume?
... I thought we were talking storage not replacing NIC cards. ... cents
at the local computer store here, so yes a retool of a NETWORK ... >

site-wide spam
filters at catherders.com. ... (SuSE)
Re: Shared storage question.
... "A SAN is also referred to as a Storage Area Network. ...

network-attached
storage server is attached to another server through the ...

(microsoft.public.windows.server.clustering)
[PAPER] Juggling with packets: floating data storage
... properties of the Internet or any other large network to create ...

volatile
distributed data storage of a large capacity. ... As a typical SMTP server

prevents
to relay mail from the client to ... (Bugtraq)
[VulnWatch] [PAPER] Juggling with packets: floating data storage
... properties of the Internet or any other large network to create ...

volatile
distributed data storage of a large capacity. ... As a typical SMTP server

prevents
to relay mail from the client to ... (VulnWatch)

-----------------------------------------------------------------------------

---

We are proud to have Web Hosting and Rack Housing from 9 Net Avenue

Deutschland.
(09)
Security UNIX Linux Coding Shop Directory Usenet

Mailing-ListsNewsgroupsAboutPrivacySearch  unix.derkeiler.com  > Newsgroups 

> comp.unix.solaris  > 2006-06      
---
Re: Optimizing memory usage of high performance webserver

-----------------------------------------------------------------------------

---

From: ptribble@xxxxxxxxxxx (Peter Tribble)
Date: 11 Jun 2006 18:32:31 GMT

-----------------------------------------------------------------------------

---
In article <1150046519.938856.195290@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
fredje110@xxxxxxxxxxx writes:

Hello!

We are running 4 Sun Fire V440 with 4 CPU's and 8 GB RAM as Apache
webservers. We do not use a dedicated caching/SQUID tier and we do not
have dedicated image servers (the amount of pictures on our website is
very large). It should not be any surprise that our internal discs have
reached their I/O limits.

Actually, this isn't obvious. Disks are fairly fast, networks
(affordable wide-area ones anyway) somewhat less so, high-traffic web
sites stress the TCP/IP stack.

On the other site the output of top shows the following:

Memory: 8192M real, 4892M free, 177M swap in use, 9538M swap free

It would make a big difference if we could load 3-4 GB of cached
content into memory (RamDisk?)

Our web tier is running Sun Solaris 8. How could we tune/convince
Solaris to load more cached content into memory and reduce disk I/O.

It does that anyway. Solaris will use all free memory as a disk
cache. So you already have a disk cache of 5G or so (based on the top
output).

Assuming that you have some basis for your 3-4G number, then that
should be the effective working set and everything should be cached in
memory already.

What do your apache traffic, network traffic, and disk I/O actually
look like? Request rate, average request size, network utilization,
disk read/writes and transfer sizes? Do you know whether it's file I/O
or directory lookups that are loading your disks?

You could try Solaris NCA, which is a direct kernel cache.

--
-Peter Tribble
L.I.S., University of Hertfordshire - http://www.herts.ac.uk/
http://www.petertribble.co.uk/ - http://ptribble.blogspot.com/
.

-----------------------------------------------------------------------------

---

References:
Optimizing memory usage of high performance webserver
From: fredje110
Prev by Date: Re: Optimizing memory usage of high performance webserver
Next by Date: Acrobat reader
Previous by thread: Re: Optimizing memory usage of high performance webserver
Next by thread: Re: Optimizing memory usage of high performance webserver
Index(es):
Date
Thread