adding this new tool from Brian Finley. He put it in trunk
git-svn-id: https://svn.code.sf.net/p/xcat/code/xcat-core/branches/2.8@15340 8638fb3e-16cb-4fca-ae20-7b5d299a9bcd
This commit is contained in:
lissav
parent
ccd2746668
commit
372a951dee
161
xCAT-server/share/xcat/tools/test_hca_state
Executable file
161
xCAT-server/share/xcat/tools/test_hca_state
Executable file
@ -0,0 +1,161 @@
|
||||
#!/bin/bash
|
||||
|
||||
#
|
||||
# 2012.08.19 Brian Elliott Finley <bfinley@us.ibm.com>
|
||||
# - created
|
||||
# 2013.02.17 Brian Elliott Finley <bfinley@us.ibm.com>
|
||||
# - added --help output
|
||||
#
|
||||
|
||||
#
|
||||
# As per Nate Rini <nate@ucar.edu>:
|
||||
#
|
||||
# The following (Red Hat) rpms include the version of pciutils (aka
|
||||
# lspci) that can read gen3 correctly.
|
||||
#
|
||||
# pciutils-3.1.10-2.el6.x86_64.rpm
|
||||
# pciutils-libs-3.1.10-2.el6.x86_64.rpm
|
||||
#
|
||||
# This may impact the "PCI:" speed test performed by this script.
|
||||
#
|
||||
|
||||
NR=$1
|
||||
|
||||
offer_help() {
|
||||
|
||||
PROGNAME=$(basename $0)
|
||||
|
||||
# At some point, source a file for this info for the whole suite
|
||||
VERSION=3.2.27
|
||||
|
||||
cat <<EOF
|
||||
$PROGNAME (part of the BEF_Scripts for xCAT) v$VERSION
|
||||
|
||||
Usage: $PROGNAME NODERANGE | xcoll
|
||||
|
||||
--help Display this help output.
|
||||
|
||||
EOF
|
||||
|
||||
cat <<'EOF'
|
||||
Purpose:
|
||||
|
||||
This tool provides a quick and easily repeatable method of
|
||||
validating key InfiniBand adapter (HCA) and node based InfiniBand
|
||||
settings across an entire cluster.
|
||||
|
||||
Having consistent OFED settings, and even HCA firmware, can be very
|
||||
important for a properly functioning InfiniBand fabric. This tool
|
||||
can help you confirm that your nodes are using the settings you
|
||||
want, and if any nodes have settings descrepancies.
|
||||
|
||||
|
||||
Example output:
|
||||
|
||||
In the output below, note the following discrepancies for the node
|
||||
storage1 (as compared with all of the other nodes): "rate",
|
||||
"PortRcvErrors", and "IB HCA FW Active". Looks like that guy may
|
||||
need a restart of the OFED stack to engage his newly installed
|
||||
firmware. ;-)
|
||||
|
||||
[root]# test_hca_state compute,storage | xcoll
|
||||
====================================
|
||||
storage1
|
||||
====================================
|
||||
OFED: MLNX_OFED_LINUX-1.5.3-4.0.22.3 (OFED-1.5.3-4.0.22):
|
||||
mlx4_0/1:
|
||||
phys state: 5: LinkUp
|
||||
rate: 40 Gb/sec (4X FDR10)
|
||||
PCI: Gen3
|
||||
SymbolErrorCounter: 0
|
||||
PortRcvErrors: 31
|
||||
IB HCA FW Installed: 2.11.500
|
||||
IB HCA FW Active: 2.10.2372
|
||||
mlx4_core set_4k_mtu: 0
|
||||
mlx4_core log_num_mtt: 20
|
||||
mlx4_core log_mtts_per_seg: 7
|
||||
IPoIB Mode: connected
|
||||
IPoIB MTU: 65520
|
||||
IPoIB recv_queue_size: 8192
|
||||
IPoIB send_queue_size: 8192
|
||||
|
||||
====================================
|
||||
compute,storage2,storage3,storage4
|
||||
====================================
|
||||
OFED: MLNX_OFED_LINUX-1.5.3-4.0.22.3 (OFED-1.5.3-4.0.22):
|
||||
mlx4_0/1:
|
||||
phys state: 5: LinkUp
|
||||
rate: 56 Gb/sec (4X FDR)
|
||||
PCI: Gen3
|
||||
SymbolErrorCounter: 0
|
||||
PortRcvErrors: 0
|
||||
IB HCA FW Installed: 2.11.500
|
||||
IB HCA FW Active: 2.11.500
|
||||
mlx4_core set_4k_mtu: 0
|
||||
mlx4_core log_num_mtt: 20
|
||||
mlx4_core log_mtts_per_seg: 7
|
||||
IPoIB Mode: connected
|
||||
IPoIB MTU: 65520
|
||||
IPoIB recv_queue_size: 8192
|
||||
IPoIB send_queue_size: 8192
|
||||
|
||||
EOF
|
||||
}
|
||||
|
||||
if [ -z $NR ]; then
|
||||
offer_help
|
||||
echo "Suggestion: Try specifying a NODERANGE"
|
||||
echo
|
||||
exit 1
|
||||
fi
|
||||
|
||||
case "$NR" in
|
||||
#
|
||||
# Match -h, -help, --help
|
||||
#
|
||||
-h*|--h*)
|
||||
offer_help
|
||||
exit 0
|
||||
;;
|
||||
esac
|
||||
|
||||
|
||||
xdsh $NR -t3 '
|
||||
echo -n "OFED: " ; ofed_info | head -n 1
|
||||
ibstatus | grep -B7 InfiniBand | egrep "(phys state|rate|device)" | perl -pi -e "s/^\s+/ /; s|Infiniband device .(.*). port (\d+).*|\$1/\$2:|;"
|
||||
HCA_PCI_ID=$(lspci | egrep "Network controller.*Mellanox" | sed "s/ .*//")
|
||||
PCI_SPEED=$(lspci -vvvxxx -s $HCA_PCI_ID | egrep "^70: " | perl -pi -e "s/^70: [0-9]+ [0-9]+ (\d+).*/\$1/")
|
||||
echo -n "PCI: "
|
||||
if [ "$PCI_SPEED" == 83 ]; then
|
||||
echo -n Gen3
|
||||
elif [ "$PCI_SPEED" == 82 ]; then
|
||||
echo -n Gen2
|
||||
else
|
||||
echo -n Dunno
|
||||
fi
|
||||
echo
|
||||
|
||||
perfquery | egrep "(PortRcvErrors|SymbolErrorCounter)" | perl -pi -e "s/\.+/ /g"
|
||||
|
||||
for HCA in $( lspci -xxxvvv | grep Net | grep Mell | sed "'s/ .*//'" )
|
||||
do
|
||||
echo -n "IB HCA FW Installed: " ; mstflint -d $HCA q | grep FW | sed "'s/.*: *//'"
|
||||
done
|
||||
for CA in $( ibstat -l )
|
||||
do
|
||||
ibstat $CA | grep -q InfiniBand && ibstat $CA | grep Firmware | sed "'s/.*: *//'" | xargs echo "IB HCA FW Active: "
|
||||
done
|
||||
|
||||
# mlx4_core
|
||||
MODULE=mlx4_core
|
||||
MLX4_CORE_PARAMS="set_4k_mtu log_num_mtt log_mtts_per_seg"
|
||||
for param in $MLX4_CORE_PARAMS
|
||||
do
|
||||
echo -n "mlx4_core $param: " ; cat /sys/module/$MODULE/parameters/$param
|
||||
done
|
||||
|
||||
echo -n "IPoIB Mode: " ; cat /sys/class/net/ib0/mode
|
||||
echo -n "IPoIB MTU: " ; cat /sys/class/net/ib0/mtu
|
||||
file=recv_queue_size ; echo -n "IPoIB $file: " ; cat /sys/module/ib_ipoib/parameters/$file
|
||||
file=send_queue_size ; echo -n "IPoIB $file: " ; cat /sys/module/ib_ipoib/parameters/$file
|
||||
' | perl -pi -e 's/(ssh: connect to host) \S+ (.*)/$1 $2/'
|
||||
Loading…
Reference in New Issue
Block a user