SCSI Config error light with a SA6402 in DL380g3 Duplex SCSI

[Bob_176]

Looks like a similar question went unanswered in an earlier post:
DL380 G3 Duplex SCSI with SA5302

Hopefully there's a answer this time around.

I've configured Duplex SCSI Drives on the DL380g3. I hook up the terminator, and attach the cables, SA6402(SmartArray) Port 1 to Drives 0,1. I then Hook up SA6402 Port 2 to Drives 3-5. When I fire up the server the SCSI Config Error Light comes on. The Duplex Light goes out, and the server appears to work fine. I'm using two of the 'Long SCSI cable's, one provided with each DL380.

The same drives connected to the embedded 5i controller operate without the SCSI config error light. Is not using the 5i causing the Error Light, if so Why?

I haven't experienced any disk errors with the SA6402, and get 240 Mbyte/sec sustained writes to a 4-disk Raid0 split across the two internal 6402 ports. I need the performance, but the maint folks can't tolerate the the SCSI error light.

Are there really SCSI termination problems, or is this a design flaw in SCSI-config-error detection circuits?

[Prashant (I am Back)]

Hi,

Check the attach image is the communication method we are using.

regards,
Prashant S

[Bob_176]

Thanks, but the image shows cable 1 attached to the 5i interface. To achieve this level of performance, I need to split the disk array across the two internal channels. Currently I have cable 1 connected to the second internal connector on the SA6402. If connected as in the image, I get same Raid0 performance as the much less expensive SA641, which only has the single internal interface.

I also want to keep a small form factor, using the internal disks rather than using an external disk chassis with the SA6402.

[Gus Kwong]

Hi Bob,

Would you let us know if there is anything appeared in the Event Viewer (I assume it's a Win2k Server), HP Array Configuration Utility's details in the "More Information", and most important of all, Insight Manager messages.

With this kind of flashing LEDs going on, there must be some kind of alerts stating the error. And we can then narrow it down from there!

Gus

[Prashant (I am Back)]

Hi,

so we are not using 5i at all, infact we are connecting both the cable to sa6402 on diff channels.

But supported option avaible are as follow:

The simplex/duplex SCSI backplane supports four cabling configurations, including:

Integrated simplex - All internal hard drives are using the embedded Smart Arrary 5i Plus controller that controls up to six hard drives through one SCSI bus. The server
ships standard with this configuration.

Integrated duplex - All internal hard drives are using the embedded Smart Array 5i Plus Controller that controls up to six hard drives through two SCSI buses:

PCI simplex - All internal hard drives are using an optional PCI array controller that controls up to six hard drives through one SCSI bus.

PCI duplex - All internal hard drives are split between two SCSI controllers. The embedded Smart Array 5i Plus Controller is connected up to two hard drives on one SCSI bus and the optional PCI array controller installed into an available PCI slot controls up to four hard drives on the other SCSI bus.

Regards,
Prashant S.

[Bob_176]

Thanks everyone for your help so far.

It seems the "PCI Duplex" option above is misnamed. It's really PCI+integrated Duplex. It looks like I'm using an unsupported configuration, a true PCI duplex configuration. By splitting the raid across the two SA6402 ports (rather than just using a single port on the SA6402) I get a significant Raid0 speed-up. I've actually tried the 4 configs listed, and the on-board red SCSI Config Error light stays off. It's this fifth config that has this irksome quality.

I'm running on Redhat Enterprise W3 udpate 1. I haven't any experience with Event Viewer (I don't think it's installed? Or Windows only?). I've looked in /var/log/messges for messages about disk errors, but nothing shows up. The cciss driver indicates everythings good. In fact, all indications are that the system is happy and healthy. The integrated 5i doesn't handle ultra320 (only SCSI-3) - perhaps the circuitry behind the light is spoofed by the midclock data transitions of the ultra320 bus?

I've attached my configuration details from cpqacuxe-7.0-1 and some entries from the log and driver status files.

[Prashant (I am Back)]

Hi

If you are using linux still u can use survey to get any error logs reported by server, that would be under /var/compaq/survey

following options can be used to see the output.

Command Line Options
-s mmhhddMMDD Schedule automatic execution. mmhhddMMDD
mm=minute, hh=hour, dd=day, MM=month,
DD=Day of Week. Use a single * for don't care
fields and all fields must be two digits (02
instead of 2). See example below.
NOTE: This command simply adds an entry to
your crontab. For more Information on crontab
type 'man crontab'.
-u Unschedule of survey running.
NOTE: This command will remove all entries in
your crontab that refer to
"/opt/compaq/survey/survey".
-c FILE1 [FILE2] Compare a file to the current output or if two
files are specified compare the files without
generating a new report.
-t Output to the terminal.
NOTE: -t and -o cannot be used together.
-o FILE Specify name of the output file. Default is
formed using the current date
at: /var/compaq/survey/surveyCCYY-MM-DD-HH-
mm-ss.txt (CC=Century,YY=Year,
MM=Month, DD=Day, HH=Hour, mm=minute,
ss=second).

regards,
Prashant S.

[Gus Kwong]

Hi Bob,

It looks like you have pretty much sorted yourself out! It was good of you to point out even with the highend SA6402 U320 RAID controller, the DL380g3 is still running on Ultra3 with its Ultra2/Ultra3 backplane --- The "Grand Champion" chipset is starting to show its age (already?!).

The Event Viewer is a Microsoft's thing --- and don't you linux-boys start laughing :)

I think it is a benefit to have Insight Manager, or HP's new SIM, running somewhere on the network --- you can see pretty much every error message if anything goes wrong! And those HP techies are beginning to ask about error messages from Insight Manager when you log a call, especially when you place pre-failure warranty call.

Gus

[Douglas F. Gibbs]

Gus seemed to think the problem was sorted out, but aside from verifying that Bob was using a non-standard configuration, the problem still exists. I, too, am having this problem trying to use a similar configuration.

Note that contrary to Gus's reply, the internal backplane is U320 "ready" according to the specs, and will do U320 on all 6 bays. The backplane is split to run the left 4 bays on one channel and the right two on the other.

Apparently it was determined that people would want to run the left four on a PCI card at U320, but leave the system drives mirrored on the embedded Ultra3 (at U160). Since I have a nice, two channel PCI card, I want both channels on U320.

Cabling as described gets that, and as near as I (we) can tell, the error lights are spurious. What is needed is either engineering confirmation that this is the case, or better yet, a solution that will make the lights happy and functional under our entirely U320 configuration.

Thanks,
DFG

[Terry Hutchings]

The only pci duplex configuration is the one shown in the illustration sent by Prashant.

The configuration you're looking for is available in the DL380 G4, but not the G3 model.