Operating System - Linux
1748092 Members
5807 Online
108758 Solutions
New Discussion юеВ

Re: SLES 9 - Hi Perf mouse blows out fiber config?????

 
SOLVED
Go to solution
Don Vanco - Linux Ninja
Regular Advisor

SLES 9 - Hi Perf mouse blows out fiber config?????

I have an HP Appliance - it is branded as an EVA Cluster File Server for Linux, but it is a DL380G5 running SLES9 and PolyServe.

I have 2 nodes in a cluster.

Because of my use of iLO and the graphical nature of the SLES tools HP provides, I enabled the HP High Performance mouse.

This resulted in the mouse going totally wonky and trashing the desktop menu, taskbars, etc.

Naturally - I de-activated the HP Hi Perf mouse.

On reboot - my fiber controllers no longer come on line. I have to manually modprobe the qla2xxx module or they are dead.

I can get no assistance from HP (have had a ticket open for a day on another issue) - no one seems to know anything about this package that HP sells for a cool $81k.

If anyone wants to address my original issue, these things come with an add-in dual port NIC. The on-boards are working OK, but the add-in cards will not pass packets. They take an IP, and the switch side shows them connected, but the queues show zero packets passed and they don't ping / etc.

11 REPLIES 11
Nuwan Alwis
Valued Contributor

Re: SLES 9 - Hi Perf mouse blows out fiber config?????

Hi Don,
Dont get panic coz DL servers number of these kind of issues. im sure your driver will be loaded from the OS because there is no way the ilo operations directly interfere OS configurations.
check /proc/scsi/scsi
messages log

following are my suggestions.
1)Do a power recycle on the server ( i mean remove the AC inputs and keep it for few mins and plug them back.
else
i dont see any other solution than unplug and replug the HBAs

once i had a same model server continuously rebooting suddenly. only thing i did was unplug and replug the HBAs. :-)

Good Luck...!
Jimmy Vance
HPE Pro
Solution

Re: SLES 9 - Hi Perf mouse blows out fiber config?????

When you installed the high performance mouse, did it rebuild the initrd image? The only way I can think of that the qlogic drivers aren't loading is they are not present in the current initrd image.

During inital boot or looking in dmesg, do you see the QLogic drivers trying to load?

In SLES the modules that will be added to the initrd image are listed in "/etc/sysconfig/kernel" under the "INITRD_MODULES=" line

If they are not listed, you can add qla2xxx_conf and qla2xxx to the line and run "mkinitrd"
No support by private messages. Please ask the forum! 
Don Vanco - Linux Ninja
Regular Advisor

Re: SLES 9 - Hi Perf mouse blows out fiber config?????

> When you installed the high performance mouse, did it rebuild the initrd image?

It did not appear to do so, but it's a moot point as HP does not put the qla2xxx in /etc/sysconfig/kernel - so it would have never been in the initrd anyway (unless manually forced in via command line at the time of making the initrd)

>you can add qla2xxx_conf and qla2xxx to the line and run "mkinitrd"

The issue here is that this is an """appliance""" from HP. I have found after a couple days on the system that there's ALL KINDS of "black magic" being used (as opposed to the logical thing, doing it the way Linux does it). I have no idea what HP did to get the drivers to be boot persistent the first time (there is a script that you run to determine the WWIDs of the attached storage, the first time you run it the drivers load, and remain persistent on reboots - this is what the mouse driver apparently blew out).

I should simply withdraw the question as no one short of an HP/PolyServe person that helped make these appliances is going to be able to help. I should have mentioned that I'm aware of how to get the drivers to load (again, as it relates to "standard" Linux practices) - I was just poking about to see if anyone ever saw a similar issue with the iLO mouse ruining their config.

I should have known something was up when the """appliance""" had execute perms TURNED OFF on the already-installed mouse driver package. Apparently it would have killed HP to go one step further and actually putting it into the manual.

Hell, I don't even know why they used SLES 9 in the first place - it's a garbage OS. The number of errors and warnings on this thing at boot time is nothing short of comical - it even runs a "release candidate" version of XFRee86. This is one sweet appliance...



And the original NIC issue? (where the add-in cards are not passing packet) HP wants to send a technician to replace both NICs with the same NICs. Brilliant. Surely both these machines shipped from the factory with bad NICs, and it's merely going to be a case of a hardware swap - Vs - IT'S AN UNSUPPORTED NIC.

2 days with an open service call, a customer that bought a million dollars of storage (just shy of a petabyte) and I can't get an HP PolyServe person to speak with me in person - I have to have them "translated" though lower level people, and participate in volleys via email where I get to spend 45 minutes typing what I could explain verbally in 5 minutes. If i get told one more time to turn of TCP offloading I'm going to crawl through the phone and kill someone. I am amazed the customer hasn't told us to take this gear out to the treelawn.
Matt Weil
Advisor

Re: SLES 9 - Hi Perf mouse blows out fiber config?????

What are the nic issues? I have had similar support issues and would love to hear more.

What version of the bnx2 driver are you using? may need to up it to version: 1.7.6b F524C3FF338F5CE8E76A837
Matt Weil
Advisor

Re: SLES 9 - Hi Perf mouse blows out fiber config?????

If you are using polyserve don't you want it to load the qla2xxx driver. you can load it by running /etc/init.d/pmxs load.

what version of polysere are you running. I would like to hear more about your experiences with this product.
Don Vanco - Linux Ninja
Regular Advisor

Re: SLES 9 - Hi Perf mouse blows out fiber config?????

The appliance comes with a CD labeled as 3.5.1 v3 - I can only assume the v3 is a delimiter for the appliance install. It looks to have been built circa Q1 2007


The NIC issues are strange indeed. The host comes with v1.7, but on one of my many attempts to fix it I did build and install the latest 1.8 drivers. No change

The issue is the on-board NICs don't connect with consistency (half or full duplex) and the add-in NICs take an IP address but pass zero packets.

Tried every permutation of hard code & aout-neg on both sides. All NICs into a Cisco 2950.

Switch shows the add-in as connected at 100 full, but ethtool says it's not connected.

HP has kind of indicated I no longer need to be involved, but they will let me know what they find....
Matt Weil
Advisor

Re: SLES 9 - Hi Perf mouse blows out fiber config?????

it is a broadcom right is this a new system g5? do you know what chip it is. did you see if there is a firmware update that is needed?

what does dmesg say?

if you pull out the add in card out?

just throwing ideas out there.
Don Vanco - Linux Ninja
Regular Advisor

Re: SLES 9 - Hi Perf mouse blows out fiber config?????

I appreciate that Matt. A for effort ;)

The add-in NICs are based on the BCM5706 chipset
The on-board NICs are based on the BCM5708 chipset

I did run the firmware update - the add-in was current but the on-board needed updating. This was done AFTER the initial discovery, so no change before/after the firmware update (or utilization of the v1.8 driver)


The BIOS sees the cards, oddly enough identifying the add-in completely but only showing the on-board as Broadcom but device unknown

The driver loads:
Broadcom NetXtreme II Gigabit Ethernet Driver bnx2 v1.7.1d (December 27, 2007)

...and the NICs are detected:
ACPI: PCI interrupt 0000:18:04.0[A] -> GSI 19 (level, low) -> IRQ 19
eth0: TCP Segmentation Offload (TSO) disabled by default
eth0: Broadcom NetXtreme II BCM5706 1000Base-T (A2) PCI-X 64-bit 100MHz found at mem fa000000, IRQ 19, node addr 00:21:5a:ed:79:b0
ACPI: PCI interrupt 0000:19:05.0[A] -> GSI 16 (level, low) -> IRQ 16
eth1: TCP Segmentation Offload (TSO) disabled by default
eth1: Broadcom NetXtreme II BCM5706 1000Base-T (A2) PCI-X 64-bit 100MHz found at mem fc000000, IRQ 16, node addr 00:21:5a:ed:79:b2
ACPI: PCI interrupt 0000:03:00.0[A] -> GSI 16 (level, low) -> IRQ 16
eth2: TCP Segmentation Offload (TSO) disabled by default
eth2: Broadcom NetXtreme II BCM5708 1000Base-T (B2) PCI-X 64-bit 133MHz found at mem f4000000, IRQ 16, node addr 00:22:64:c1:cb:8e
ACPI: PCI interrupt 0000:05:00.0[A] -> GSI 17 (level, low) -> IRQ 17
eth3: TCP Segmentation Offload (TSO) disabled by default
eth3: Broadcom NetXtreme II BCM5708 1000Base-T (B2) PCI-X 64-bit 133MHz found at mem f6000000, IRQ 17, node addr 00:22:64:c1:cb:8c

However, when the driver loads, it never seems to "see" the add-in NIC as it should:
(note the absence of eth0 and eth1)
bnx2: eth2 NIC Copper Link is Down
bnx2: eth3 NIC Copper Link is Down
bnx2: eth2 NIC Copper Link is Up, 100 Mbps full duplex
bnx2: eth3 NIC Copper Link is Up, 100 Mbps half duplex


I did not pull the card - with this being an "appliance" I did not want to crack the lid.

This is also the reason I did not upgrade the HAB firmware - I looked at the update and it was really only pertinent to the Blade HBAs, and given the tie between driver and firmware I was not willing to change the config as shipped from HP unless I saw issues. While I am seeing an issue, it is purely OS/software related.

Jimmy - I was not aware PolyServe configs had a separate locale for driver loads - this is probably a good tip, but alas I no longer have access to these hosts to check the diff between the one that works and the one that does not. I will see if the customer can send the files to me. Thanks!


At the risk of waxing poetic - I really like the Linux forum here - it's literally the only ITRC forum where people answer my questions (or possibly tolerate my stupidity?) :D
Rob Leadbeater
Honored Contributor

Re: SLES 9 - Hi Perf mouse blows out fiber config?????

Hi,

Going back to your original issue, I can remember doing a HP update on a RHEL box in the past which managed to lose my connection to an EVA. It's a while ago now, but I seem to recall having to hack around with modules and kernels to get back to the original RedHat rather than HP QLA modules in order to see the SAN.

I eventually figured out that the kernel build that the HP package did (it was a ProLiant Support Pack install IIRC) had thrown up errors, but continued to build a duff kernel.

The strange thing was that the same install on an identically built box worked fine...

I stopped trusting any HP package that did a kernel build after that !!

As to your NIC issue, I wonder if you're in a similar issue, with conflicts between the bnx and e1000 drivers ?

Also, I know you've said you've tried every combination of speed/duplex, but I would sanity check that. On modern Gb NICs leaving everything at auto sense is the way to go...

Hope this helps,

Regards,

Rob