Monitoring Exadata Storage Cells - Oracle Exadata Recipes: A Problem-Solution Approach - page 398

13-11. Detecting Host Interconnect Bottlenecks
Problem
You wish to diagnose host interconnect performance issues or bottlenecks.
Solution
In this recipe, you will learn how to use HOST_INTERCONNECT metrics to identify network bottlenecks on your Exadata
storage grid. Execute the script in Listing 13-8, which will display network metrics for network drops, RDMA network
drops, and resent/RDMA retry operations in units of MB per second. If the script displays output, it is an indication
that you have a network issue of some sort.
Listing 13-8. lst13-08-interconnect-probs.sh
#!/bin/sh
# Name: lst13-09-interconnect-probs.sh
# Purpose: Show host interconnect metrics above threshold
# Usage: ./lst13-09-interconnect-probs.sh [metriccurrent|metrichistory]
SCOPE=$1 # run with either metriccurrent or metrichistory
# Adjust the following to suit your needs
CELLGROUP=/home/oracle/cell_group
MBMIN=0 ## 0 MB/second for each type of dtop
########################################
## Metrics list per type #####################################################
MBPS="N_MB_DROP_SEC N_MB_RDMA_DROP_SEC N_MB_RESENT_SEC"
##############################################################################
# dcli/cellcli commands
for met in $MBPS
do
echo "Listing ${SCOPE} for ${met} when IO rate > ${MBMIN} MB/second"
dcli -g ${CELLGROUP} "cellcli -e list ${SCOPE} \
where objectType=\'HOST_INTERCONNECT\' and name=\'${met}\' \
and metricValue \> ${MBMIN}"
done
[oracle@cm01dbm01 source]$ ./lst13-07-interconnect-probs.sh metriccurrent
Listing metriccurrent for N_MB_DROP_SEC when IO rate > 0 MB/second
Listing metriccurrent for N_MB_RDMA_DROP_SEC when IO rate > 0 MB/second
Listing metriccurrent for N_MB_RESENT_SEC when IO rate > 0 MB/second
[oracle@cm01dbm01 source]$
In Listing 13-8, we are reporting on HOST_INTERCONNECT metrics in which the network drops and retransmission
rates are above 0 MB per second; ideally, your machine should not experience network drops. These thresholds are
defined at the beginning of the script—you can change these values to report on different values as needed.