Cisco Cisco Unified Customer Voice Portal 11.0(1)

Page of 102
Failure and Restoration
If the Reporting Server fails, messages destined for the Reporting Server are buffered by the
Call Server, in memory, up to 200,000 messages. After that limit is reached, all new messages
are dropped.
If the database connection fails, the Reporting Server sends out an SNMP alert and starts
persisting messages to a file, up to a user specified limit. During this time the Reporting
Server stays In Service. When 75% of the specified limit is reached, a warning is written to
the log file. Once 100% of the limit is reached, an SNMP alert is sent out and the Reporting
Server goes into Partial Service—any new messages may be dropped.
When the database connection comes back up, the Reporting Server goes into recovery mode
and changes its state to Partial Service if it is not in that state already. It then starts reading
messages from the file and committing them to the database. Depending on the size of the
file, it may take a long time (sometimes hours) to commit all of the data to the database. Any
new messages that come in during recovery will be buffered in memory. There is, however,
a limit to the number of messages that the Reporting Server can buffer. This is true regardless
of the mode or state it is in. When the number of buffered messages reaches 100,000, an
SNMP alert is sent out to warn the user. At 200,000 another SNMP alert is sent out and all
new messages' detail information is dropped—keeping only basic data like call, call event,
and session information. Also at 200,000, the Reporting Server changes its state to Partial
Service, if it is not already in that state. After the total number of buffered messages reaches
300,000, another SNMP alert is sent out and all new messages are dropped from that point
forward.
When the number of messages in memory drops back below 50,000, an SNMP alert is sent
out stating that the queue size is back to normal, and the Reporting Server’s state goes back
to In Service.
If, on startup, a persistent file exists, the Reporting Server stays in Partial Service and goes
into recovery mode as described above.
During a database purge operation, the Reporting Server disconnects from the database and
starts buffering messages in memory until the purge is done. The same memory limitations
as described above apply in this case as well.
Caution: When the Reporting Server is in Partial Service, there are no guarantees that
new messages will be kept and committed to the database. They will be buffered in memory
for as much as possible, but at some point they may be dropped either partially or fully.
Reporting Guide for Cisco Unified Customer Voice Portal Release 7.0(1)
20
Chapter 3: - Managing the Database
Failure and Restoration