SPECvirt ® Datacenter 2021 Technical Support

Use this document to troubleshoot and resolve errors you might encounter.

On the svdc-director VM

Setting the MAC address on vSphere

Some versions of vCenter Server 8 might not save your changes when you try to change the svdc-director VM’s MAC address. To work around this limitation:

  1. Determine the host on which the svdc-director VM resides and log into the ESXi host’s graphical user interface

  2. Select the svdc-director VM then Edit Settings and expand Network adapter 1 (private network)

  3. Under MAC Address:

  1. Change the assignment from Automatic to Manual

  2. Enter the required MAC address based on your prefix in Control.config (default is 42:44:49:00:00:02)

  1. Log out of the ESXi host’s interface. In the vSphere client, check that the MAC address is correct

  2. Resume the installation steps in the SPECvirt Datacenter 2021 User Guide.

Submitting a measurement manually

You can submit a measurement manually:

./submit.sh -c ../config/cloudpump.xml -w ../config/workloads/specvirt/workload.xml -d ../config/workloads/specvirt/deployment.xml -r ../config/workloads/specvirt/run.xml

Encountering a Java connection error

If you receive an error message:

Failed to connect to master svdc-director:19998
java.rmi.ConnectException: Connection refused to host: svdc-director; nested exception is: java.net.ConnectException: Connection refused (Connection refused)

Run $CP_BIN/resetprimes.sh.

Setting up SSH and vSphere

vSphere 7.0.3 and later only supports ECDSA with SSH. To add ECDSA support between the svdc-director VM and the vSphere hosts, on the svdc-director VM issue the following:

ssh-keygen -t ecdsa -f /root/.ssh/id_ecdsa -N ''
for Host in `grep Host /etc/hosts | awk '{print $1}'`
do
    cat ~/.ssh/id_ecdsa.pub | ssh $Host 'cat >> /etc/ssh/keys-root/authorized_keys'
    ssh $Host 'chmod +t /etc/ssh/keys-root/authorized_keys'
    ssh $Host '/etc/init.d/SSH restart'
done

Encountering issues with the reporter

The reporter requires requires a space after each blank line with an “= ” in Control.config. A missing space after each blank line with an “= ” results in the following error:

                      SPECvirt® Datacenter 2021 Reporter 1.0
 Copyright (c) 2010-2022 Standard Performance Evaluation Corporation
 --------------------------------------------------------------------

 > Raw Result File - specvirt-datacenter-2022-07-07_14-31-43.raw

 > Generating Reports in HTML Format

 > Verifying Raw Result File..

 > Renaming Raw Result File: specvirt-datacenter-2022-07-07_14-31-43.raw.backup

 > Loading Raw Result File..

 > Recreating Raw Result File: specvirt-datacenter-2022-07-07_14-31-43.raw

Exception in thread "main" java.lang.ArrayIndexOutOfBoundsException: 1
      at org.spec.virt.reporter.Results.loadLinkedHashmapConfig(Results.java:340)
      at org.spec.virt.reporter.Results.rewriteResFile(Results.java:171)
      at org.spec.virt.reporter.Reporter.main(Reporter.java:166)

To fix this error, comment out the Control.config parameter “virtCert = “:

# virtCert =

On client VMs

For problems deploying clients, check the following.

  • Ensure that the svdc-template VM’s and svdc-director VM’s virtual NIC has assigned the correct MAC address. Check the svdc-template’s eth0 vNIC settings to ensure that the MAC prefix (the first three octets) assigned to the vNIC matches the value for MACAddressPrefix that you set in Control.config for the testbed network. The default for svdc-template is 42:44:49:00:00:00. The default for the svdc-director VM is 42:44:49:00:00:02.

  • Ensure that the svdc-director VM’s virtual NIC has assigned the correct IP address. Check that the svdc-director VM’s IP address uses the value of IPAddressPrefix and 0.2. The default IP address for the svdc-director VM is 172.23.0.2.

  • Use ping to make sure that the svdc-director VM can contact the management server.

Submissions

If you encounter problems during the collection of the support tarball archive at the end of a measurement, you can rerun ${virtVendor}/collectSupport.sh manually. The support archive is valid as long as you haven’t made changes to the testbed.

HammerDB

HammerDB creation takes a long time

Running prepSUT.sh takes a long time and appears to hang. CPU utilization on the svdc-director VM is idle, and the console output reads:

...
ALL VIRTUAL USERS COMPLETE
hammerdb>DB load successful. Creating DB backup. **Note, this takes several minutes (or more).
Please be patient...

This pause occurs as the harness backs up the HammerDB database on the svdc-director VM and could take several hours.

Monitoring HammerDB transactions

To monitor active HammerDB transactions during a measurement:

  1. Get the jobid from http://svdc-txxx-happ:8080/jobs

  2. Enter the jobid at http://svdc-txxx-happ:8080/jobs?jobid=[jobid] and refresh browser to see updates

If HammerDB reports no transactions, check if:

  • svdc-t*-hdb doesn’t have /opt/hammerdb_mysql_backup/hammerDB-backup.tgz with size = 13281381806

  • svdc-t*-happ /opt/HammerDB/*tcl.orig doesn’t contain svdc-t*-hdb and instead contains TARGETHOST

If either of these conditions exist, prepSUT.sh failed. Rerun $CP_COMMON/prepSUT.sh, and tail -f $CP_COMMON/prepSUT.log for output.

BigBench

Accessing the BigBench console

To access the BigBench web console and monitor transaction job history:

http://svdc-txxx-bbnn:19888/jobhistory/app - shows history of 38 jobs completed during initial load as well as all past jobs

http://svdc-txxx-bbnn:8088/cluster/apps - monitor active BigBench jobs

http://svdc-txxx-bbnn:8088/cluster/nodes - monitor BigBench data node status

Checking BigBench status

To check the status of BigBench services:

ssh svdc-txxx-bbnn "/opt/createApacheCluster/checkHadoopStatus.sh"

Copyright 2021 Standard Performance Evaluation Corporation (SPEC). All rights reserved.