I've noticed an issue with my e5K. I rebooted yesterday (I rarely
reboot this box since you can do a lot dynamically) and none of my
zones were coming up. So I checked and the disk on which I have zones
was not mounted.
, not a problem, I issue a mount and start my zones. Didn't think
too much of it at the time.
This morning I was loading a database on a different drive (that the
database zone has access to) and noticed how slow it was.
iostat showed 1100Kw/s and %b 96 to 100!
Isn't that kind of slow for the SCSI on the IBoard to a ST130 with 10
K rpm U320 drives (I know it only does u160 but still). Digging in
dmesg I see:
Nov 23 12:44:41 e5000 scsi: [ID 365881 kern.info] fas: 4.0: cdb=[ 0x12 0x0
0x0
0x0 0x30 0x0 ]
Nov 23 12:44:41 e5000 scsi: [ID 107833 kern.warning] WARNING:
/sbus@3,0/SUNW,fas
@3,8800000 (fas0):
Nov 23 12:44:41 e5000 Connected command timeout for Target 4.0
Nov 23 12:44:41 e5000 scsi: [ID 107833 kern.warning] WARNING:
/sbus@3,0/SUNW,fas
@3,8800000 (fas0):
Nov 23 12:44:41 e5000 Target 4 reducing sync. transfer rate
Nov 23 12:44:41 e5000 scsi: [ID 107833 kern.warning] WARNING:
/sbus@3,0/SUNW,fas
@3,8800000/sd@2,0 (sd2):
Nov 23 12:44:41 e5000 SCSI transport failed: reason 'reset': retrying
command
Nov 23 12:44:44 e5000 scsi: [ID 107833 kern.warning] WARNING:
/sbus@3,0/SUNW,fas
@3,8800000/sd@2,0 (sd2):
Nov 23 12:44:44 e5000 Error for Command: write(10) Error
Level:
Retryable
Nov 23 12:44:44 e5000 scsi: [ID 107833 kern.notice] Requested Block:
2128694
Error Block: 2128694
Nov 23 12:44:44 e5000 scsi: [ID 107833 kern.notice] Vendor: SEAGATE
Serial Number: 0043K81286
Nov 23 12:44:44 e5000 scsi: [ID 107833 kern.notice] Sense Key: Unit
Attentio
n
Nov 23 12:44:44 e5000 scsi: [ID 107833 kern.notice] ASC: 0x29 (scsi bus
rese
t occurred), ASCQ: 0x2, FRU: 0x2
Nov 23 12:44:44 e5000 scsi: [ID 193665 kern.info] sd4 at fas0: target 4 lun 0
I also have a CDRM and DDS tape on the same controller (connected
thru the front and into the 5000 chassis.
Any idea what is going on?
Francois
rescue list -

24/11/05, Francois Dion <francois.dion (AT) gmail (DOT) comwrote:
I've noticed an issue with my e5K. I rebooted yesterday (I rarely
reboot this box since you can do a lot dynamically) and none of my
zones were coming up. So I checked and the disk on which I have zones
was not mounted.

, not a problem, I issue a mount and start my zones. Didn't think
too much of it at the time.

This morning I was loading a database on a different drive (that the
database zone has access to) and noticed how slow it was.

iostat showed 1100Kw/s and %b 96 to 100!

Isn't that kind of slow for the SCSI on the IBoard to a ST130 with 10
K rpm U320 drives (I know it only does u160 but still). Digging in
dmesg I see:

Nov 23 12:44:41 e5000 scsi: [ID 365881 kern.info] fas: 4.0: cdb=[ 0x12 0x0
0x0
0x0 0x30 0x0 ]
Nov 23 12:44:41 e5000 scsi: [ID 107833 kern.warning] WARNING:
/sbus@3,0/SUNW,fas
@3,8800000 (fas0):
Nov 23 12:44:41 e5000 Connected command timeout for Target 4.0
Nov 23 12:44:41 e5000 scsi: [ID 107833 kern.warning] WARNING:
/sbus@3,0/SUNW,fas
@3,8800000 (fas0):
Nov 23 12:44:41 e5000 Target 4 reducing sync. transfer rate
Nov 23 12:44:41 e5000 scsi: [ID 107833 kern.warning] WARNING:
/sbus@3,0/SUNW,fas
@3,8800000/sd@2,0 (sd2):
Nov 23 12:44:41 e5000 SCSI transport failed: reason 'reset': retrying
command
Nov 23 12:44:44 e5000 scsi: [ID 107833 kern.warning] WARNING:
/sbus@3,0/SUNW,fas
@3,8800000/sd@2,0 (sd2):
Nov 23 12:44:44 e5000 Error for Command: write(10) Error
Level:
Retryable
Nov 23 12:44:44 e5000 scsi: [ID 107833 kern.notice] Requested Block:
2128694
Error Block: 2128694
Nov 23 12:44:44 e5000 scsi: [ID 107833 kern.notice] Vendor: SEAGATE
Serial Number: 0043K81286
Nov 23 12:44:44 e5000 scsi: [ID 107833 kern.notice] Sense Key: Unit
Attentio
n
Nov 23 12:44:44 e5000 scsi: [ID 107833 kern.notice] ASC: 0x29 (scsi bus
rese
t occurred), ASCQ: 0x2, FRU: 0x2
Nov 23 12:44:44 e5000 scsi: [ID 193665 kern.info] sd4 at fas0: target 4 lun
0
--
I also have a CDRM and DDS tape on the same controller (connected
thru the front and into the 5000 chassis.

Any idea what is going on?

Francois

rescue list -

It might not be the disk, have you changed the termination. I have had
similar errors (although without the write error) which has been
termination.

Dan

rescue list -

11/24/05, Dan Williams <williams.dan (AT) gmail (DOT) comwrote:
It might not be the disk, have you changed the termination. I have had
similar errors (although without the write error) which has been
termination.

We have also seen that type of behavior on SCSI with a "marginal"
SCSI controller (replaced), and on FC with a bent-too-sharply
FC cable.

If the errors are isolated to a single disk, though, I'd try removing it
first before fiddling with other bits. I hope you have some kind of
LVM RAID at least?

=Nadine=

rescue list -

11/24/05, Gavin Hubbard <gavin.hubbard (AT) parallix (DOT) comwrote:
Hi Francois

The integrated SCSI controller on the I Board is single-ended UW-SCSI
(maximum 40MB/s) - not U160.

40MB/s would be great. But I was getting 1MB/s :(

I pulled the drive and transactions are now up to a max of about
18.5MB writing on that process

What SCSI SBUS cards offer 40MB/s?

Francois

rescue list -

11/24/05, Dan Williams <williams.dan (AT) gmail (DOT) comwrote:
It might not be the disk, have you changed the termination. I have had
similar errors (although without the write error) which has been
termination.

I was wondering about this. Altough performance was restored by
removing the disk, and putting a new disk appeared to be ok, I noticed
again subpar performance later on. I cant tell if the SCSI bus was
degraded or not, but iostat is again showing me 90-100% with a meager
1MB/s. I'm ok with 18MB/s, altough I'd want better, but 1MB/s I cannot
live with that!

I'm wondering if the on board SCSI controller cant deal with the 3
drives (id 2,3,4) in the st d130 PLUS thru the front scsi tray
adapter, the CDRM and DDS-3. Can you *not* use the back SCSI port on
thesbus/ I board on slot 0 if you use the scsi tray/adapter also?

Francois

rescue list -

11/25/05, velociraptor <velociraptor (AT) gmail (DOT) comwrote:
11/24/05, Dan Williams <williams.dan (AT) gmail (DOT) comwrote:
It might not be the disk, have you changed the termination. I have had
similar errors (although without the write error) which has been
termination.

We have also seen that type of behavior on SCSI with a "marginal"
SCSI controller (replaced), and on FC with a bent-too-sharply
FC cable.

I got some SBUS SCSI controller I can try, but I might as well update
my SBUS I board (SC, not SC+) and see if that makes any difference.

If the errors are isolated to a single disk, though, I'd try removing it
first before fiddling with other bits.

I removed it and got good performance for a while. I added another
36gb disk and for a while I had decent performance, but again, it is
slow.

I hope you have some kind of
LVM RAID at least?

Bah, no raid. I dont have enough disks to mirror anything but the boot
disk. I can restore from UFSDUMPs pretty quickly tough. If it was a
production box, I would have raid 10, but this is one of the servers
in my basement. I'm just playing with this right now. Eventually I'll
add an A5200 or something like that.

course, it complicates things that this is mounted thru a zone. I
might shutdown the zone and mount it on the global zone and see what
happens there.

Is there a script or app somewhere that has a standard throuput test?
I know about iozone but I have nothing to compare against.

Ciao,
Francois

rescue list -

Everything is "untested but looks K" unless otherwise noted. More to come
later this month!

B.

2 @ Spyrus MC Disk SCSI PCMCIA readers, in original boxes with manual,
RS/6000 AIX 3.2.5 and 4.1.1 drivers, power supply, and reader. Mostly
intended for memory cards ("SRAM, RM, TP, E(E)PRM, FLASH, and PC-ATA
disks (mechanical/silicon)"), but also supports "specific I/ cards". HD50
SCSI connectors, two PCMCIA slots. These were part of my long-ago plot to
get solid-state SCSI storage; unfortunately, while they would read and
write just fine, I couldn't quite get a Sun to boot from them. $5 each
plus shipping.

2 @ SIIG US2225 USB 1.1 Cardbus cards, HCI. New in sealed boxes. Note
that like most USB cards, these require newer slots which support a full
range of voltages -- if your laptop doesn't have a USB port, it probably
doesn't support these cards either. $5 each plus shipping.

Addonics IDE Ultra DigiDrive, new in sealed box. Mounts in a 3.5" drive
bay, one PCMCIA slot for mounting apparently any card that exposes an ATA
interface to your IDE chain. Bootable, but only if your BIS supports
booting from a Zip drive. Yet another failed experiment in solid-state
storage. $5 plus shipping.

HP "Accy CD Kit L.12.05", P/N 5064-6675. Two CDs, still shrinkwrapped.
is "HP NetServer Documentation/including HP NetServer Information
Assistant", D.12.00, document ID 7737, copyright 1993-1998, P/N 5011-7024,
for the NetServer E Series, L Series, and Retired Systems; the other is
similar but hidden by the exterior label, P/N 5011-5074, whatever it is,
it's for the NetServer LXr Pro8, LXr Pro, LX Pro, LXe Pro, LH II, LH Pro,
LD Pro, and LC II. Free for shipping.

HP document, NetServer "LH 3/LH 3r Processor Upgrade Installation Guide",
P/N 5967-2106, June 1998. Free for shipping.

HP document, NetServer "LC 3 Processor Upgrade Installation Guide", P/N
5967-5254, July 1998. Free for shipping.

Sun 530-1857-03 Rev 51, 5M HD26 to Centronics parallel cable for SS10/SS20
and various Netras. $2 plus shipping.

Wangtek 5099EN24 QIC-24 tape drive, in mounting bracket for old Sun-3
shoeboxes, with 370-1235-01 Emulex MT02 tape controller and cable from
controller to drive. Free for shipping.

Xircom PE3-10BT parallel to Ethernet 10baseT gizmo, with keyboard
passthrough power cable. Xircom never made 32-bit Windows drivers for this
unit, and the 16-bit drivers which their web page claimed were
Win98-compatible didn't work. If you have an old, old laptop which is
still running 16-bit Windows, this might just get you online, though. Free
for shipping.

2 @ Lexmark S/SC Coax/Twinax Adapter, FRU P/N 99A0629. DE9
connector, Intel 80960 onboard. No cables. $5 each plus shipping.

Sun 501-2592 sound module for SS4. I didn't look at this carefully enough
when I bought it, I thought it was an SBus sound card! $1 plus shipping.

HP RG5-4198 "Controller Board for 2000 Sheet Paper Tray for LaserJet 5Si",
$5 plus shipping.

DEC DEFEA EISA CDDI controller, dual attach, in original box, static bag
still sealed, with sealed manual and driver disks. Free for shipping.

DEC DEFEA-UA CDDI controller, single attach, in original box with manual.
Free for shipping.

12 @ 30-pin 1M 80ns SIP memory, various brands. Free for shipping.

2 @ mystery 144-pin SDIMMs from Simple Tech. Each sports eight
M5M4V17805CTP RAM chips and nothing like a Simple Tech part number. Free
for shipping.

Adtron "Solidstate Data Drive" ISA-PCMCIA interface. 16-bit ISA card with
VADEM VG-365 chip, two PCMCIA slots in 3.5" bay mounting, all four ribbon
cables. Supported by Linux, maybe NetBSD? Free for shipping.

2 @ Sysgen SC4000 SCSI/QIC-02 controller for QIC-11 drives, from Sun 2/120
units! with both SCSI and tape ribbon cables, the other with no
cables. $5 each plus shipping.

HIcard2 by RYBS Electronics, 1988. 8-bit ISA card, 256K of memory which
can boost your XT to the 640K limit and fill in memory gaps beyond!
disks, manuals. It was really cool when I still had an XT. Free
for shipping, last known working.

3Com FDDILink 3C771 EISA SAS FDDI card with MIC connector. Free for shipping.

Adaptec AHA-2842A VESA Local Bus SCSI controller with floppy controller.
HD50 external SCSI, 50-pin internal SCSI. Free for shipping.

Apple Token Ring 4/16 NUBus card, actual Apple-branded product. DE9
connector for STP. Free for shipping.

IBM 16/5 Token-Ring PCI Management Adapter, FRU 35P5409, STP (DE9) and UTP
(RJ45). Free for shipping.

HP VRM, P/N 0950-2848, for LCII/LC2, LHII, LH3, LPr. Free for shipping.

3Com 3C597-TX EISA 10/100baseT card. This is the one that can be hacked
into working in an Indigo 2. $5 plus shipping.

HP C2071T XI ethernet interface card, for LJII. You can see photos of
this card and get the software that I wrote for it at
The LJII referenced is
long gone, so this card and its manuals are now free for shipping. Last
known working.

Matrox G450 16M dual-head PCI video card, fully supported as a secondary
display by Windows. Works. $10 plus shipping.

Roland LAPC-I, full-length 8-bit ISA MIDI/synth card, one of the earliest
sound cards. Free for shipping.

rescue list -

Mon, 28 Nov 2005 11:34:44 -0500, velociraptor wrote:
11/25/05, Francois Dion <francois.dion (AT) gmail (DOT) comwrote:

I got some SBUS SCSI controller I can try, but I might as well
update my SBUS I board (SC, not SC+) and see if that makes any
difference.

We never used those for anything other than the system boot disks,
so I can't offer any suggestions in that regard.

As it happens I have an E4000 which should be similar to the E5000 in
this regard. I've had a problem with *one* disk producing the kind of
messages you were seeing (sorry not been paying attention), but
nothing since.

All my storage is attached through the I board, although it isn't being
driven very hard.

Is there a script or app somewhere that has a standard throuput
test? I know about iozone but I have nothing to compare against.

Bonnie? None of them have baselines, though as far as I know--
Bonnie++ or Postmark (which isn't really a throughput test). There's
also FileBench but that might be overkill here.

I can run both on my E4000 probably on a spare 4Gb disk, but not in
single-user mode.

rescue list -

11/25/05, Francois Dion <francois.dion (AT) gmail (DOT) comwrote:

I got some SBUS SCSI controller I can try, but I might as well update
my SBUS I board (SC, not SC+) and see if that makes any difference.

We never used those for anything other than the system boot disks,
so I can't offer any suggestions in that regard.

course, it complicates things that this is mounted thru a zone. I
might shutdown the zone and mount it on the global zone and see what
happens there.

I'd be interested to hear if that is an issue. I know there are ways
with zones to do resource management this has been
inadvertently set to low I/ resources? (stab in the dark)

Is there a script or app somewhere that has a standard throuput test?
I know about iozone but I have nothing to compare against.

Bonnie? None of them have baselines, though as far as I know--
they are all comparative. You might google to see if anyone has
pub'd "stock" results for Sun boxes, but even then, it depends on
how well they've documented their data tests as to whether it will
mean anything comparatively.

=Nadine=

rescue list -

11/28/05, Mike Meredith <very (AT) zonky (DOT) orgwrote:
Mon, 28 Nov 2005 11:34:44 -0500, velociraptor wrote:
11/25/05, Francois Dion <francois.dion (AT) gmail (DOT) comwrote:

I got some SBUS SCSI controller I can try, but I might as well
update my SBUS I board (SC, not SC+) and see if that makes any
difference.

We never used those for anything other than the system boot disks,
so I can't offer any suggestions in that regard.

As it happens I have an E4000 which should be similar to the E5000 in
this regard. I've had a problem with *one* disk producing the kind of
messages you were seeing (sorry not been paying attention), but
nothing since.

I have not gotten the issue since either, kinda weird

Bonnie++ or Postmark (which isn't really a throughput test). There's
also FileBench but that might be overkill here.

I can run both on my E4000 probably on a spare 4Gb disk, but not in
single-user mode.

My e5000 always has activity on it also. Still, I'd be curious to see
if i'm getting in the ballpark. I know what software mirroring
improves read performance quite a bit on iozone.

I've been thinking more and more to go to ZFS and compression. It
helps on lower bandwidth SCSI from my quick testing. When is ZFS going
to be available as an update on Sol 10 (if ever?). Solaris is cool
but I need a stable e5000. I have an Ultra 2 and V20z for

Anyway, back to Bonnie++ I'll install it and see what kind of results
I get. Let me know also what you get.

Francois

rescue list -