nx_nic issues

[David Livingstone]

We have two DL380G5 servers both running RHEL5(latest) connected back-to-back over their NC510C PCIe 10 gigabit nic(one per machine).
We are also running the latest PSP 8.15 on
both. Back in Dec we noted that sometimes
the link would hang so we updated to nx_nic-3.4.337-1 from 336 and the problems decreased.
Lately we have still noted some issues so
we downloaded and installed the latest packages on one machine:
- nx_nic-4.0.230-1.src.rpm
- nx_lsa-4.0.225-1.src.rpm
- nx_tools-4.0.230-1.noarch.rpm

The link between the machines is currently
running however :
1. The nx_tools pkg does not have the
4.0.230 flash image. When you flash
the card with nxflash it loads
3.4.337.
2. After upgrade the new nx_nic, nx_lsa, and
nx_intercept are loaded however
/proc/net/nx_nic/lsa_1/stats is
missing and nxoffload has no affect.

We also see the following on both servers
(one with 3.3.337 and one with 4.0.230)

022 pts/2 Sl 2:24 cmanicd
959 pts/2 S 0:00 \_ cmanicd
960 pts/2 S 0:00 \_ sh -c /opt/hp/hp-snmp-agents/nic/bin/hpetfe -A -d /var/spool/compaq/nic/nicinfo 2>/dev/null
961 pts/2 S 0:00 \_ /bin/bash /opt/hp/hp-snmp-agents/nic/bin/hpetfe -A -d /var/spool/compaq/nic/nicinfo
1491 pts/2 S 0:00 \_ /bin/bash /opt/hp/hp-snmp-agents/nic/bin/hpetfe -A -d /var/spool/compaq/nic/nicinfo
1492 pts/2 S 340:37 \_ lspci -s 0000 0b 00 0 -n -v
1493 pts/2 R 2167:39 \_ egrep -i memory at
1494 pts/2 S 0:00 \_ head -1
1495 pts/2 S 0:00 \_ awk {print $3}
1496 pts/2 S 0:00 \_ sed s/^[ ?]*//g

The lspci location is that for the 10 gigabit
board. This happens on both servers and as you can see will run forever and hang the 2381 interface to the machine.

[Steven E. Protter]

Shalom,

Ah to be on the bleeding edge.

Fun fun fun.

Are these NIC cards standard with the server?

If not, go to the manufacturer for the driver, its not in RHEL.

If they are standard with the server, then HP should provide the drivers.

Perhaps with the PSP pack that ships with the server.

SEP

[HPUX Admins]

Not much help, but I can confirm the 3.4.337
firmware is what gets loaded when using
nx_tools-4.0.230-1.noarch.rpm.

I'm bring 4 c7000 frames online with BL465c G5 blades. We're looking to run Red Hat 4.7. I have installed the latest Firware Deployment Tools 1.41 onto all the blades, and the c7000 is running Onboard Administrator 2.40.

I haven't messed with the BL-c 10Gb switch in
the frame. So far, I have installed RH 4.7 on
the blades in slots 1 and 2, but have so far
been unable to get them to talk to each other.
I have the latest nx_nic driver, link/speed/duplex look ok, but pinging from one blade to the other, I get "Destination host unreachable". I haven't done any VLAN config on the hosts nor adjusted anything on the BL-c. Looking at the port stats, it seems that I'm seeing TX packets, but no RX packets.

I opened a case yesterday...will advise if/when I hear anything.

[David Livingstone]

I also opened a case on Friday and will post the resolution.

[Mark Frenette]

Was there any resolution to this? I am having the exact same problem.

[David Livingstone]

I have had some email correspondence with tech support however we have not yet had a resolution. Supposedly other people have been having the problem and engineering is working on it.So far on the driver issues
their recommendation is to go back to 3.4.337-1. I have looked at the new psp-8.20 and unfortunately it still has the identical nx_nic-4.0.230-1.src.rpm and nx_lsa-4.0.225-1.src.rpm packages.

[David Livingstone]

While awaiting updated nx and integrated psp have downgraded the psp to 8.11 and set the nx_lsa and nx_nic drivers to 3.4.337. So far no problems.