Error starting the secondary SS - CsCorbaMgr Exception CosNaming NamingContext InvalidName

Products

• Network Observability
• Spectrum

Issue/Introduction

DX NetOps Spectrum generates the following error when starting the SpectroSERVER:

ERROR TRACE at CsCorbaMgr.cc(1271): Exception: ::CosNaming::NamingContext::InvalidName

This occurs only when starting the secondary/fault tolerant server.  The primary SpectroSERVER does not have any issue.

The $SPECROOT/SS/VNM.OUT shows:

Jan 17 03:50:19 ERROR TRACE at CsCorbaMgr.cc(1271): Exception: CORBA::TIMEOUT

Minor: 1447165953 

Completion Status: NO

 

The SpectroSERVER hangs and does not actually start.

 

Environment

This was only seen on the secondary SpectroSERVER of a Fault Tolerant pair. 

Cause

The service account running processd is not able to authenticate/register before processd starts.

Resolution

The resolution is to edit the  $SPECROOT/lib/SDPM/processd_init.sh file and add a sleep statement to allow the service account time to register.

In this section:

case "$1" in
start)
    echo "Starting SPECTRUM Process daemon."
    if [ -d /var/lock/subsys ]
    then
        touch /var/lock/subsys/processd

Add a sleep 60 statement so it will look like this:

case "$1" in
start)
    echo "Starting SPECTRUM Process daemon."
 sleep 60
    if [ -d /var/lock/subsys ]
    then
        touch /var/lock/subsys/processd

Then proceed to recover the SS:

Kill the SpectroSERVER process.

Stop processd

Start processd (it will now take longer due to the addition of the "sleep 60" parameter in the startup script)

Run OnLineBackup on the primary SS to sync the SSdb with the secondary (Ensure the $SPECROOT/SS/.VNMDBLOCK file is not present if the SpectroSERVER is stopped). Or manually copy over the backup file to the secondary SS and restore it.

Start the secondary SS