Cisco MCS-7825-H3-IPC1, Руководство по обслуживанию

Страница: 121 / 974

3-21
Cisco Unified Communications Manager Managed Services Guide
OL-22523-01
Chapter 3 Managing and Monitoring the Health of Cisco Unified Communications Manager Systems
RTMT Monitoring of Cisco Unified CM System Health
The MCS-7845 server has more processors and a lower threshold for CPU usage because the ccm
process is a multithreaded application. But the main router thread does the bulk of call processing. A
single thread can run only on one processor at any given time even when there are multiple processors
available. That means ccm main router thread can run out of CPU resource even when there are idle
processors.
With hyper-threading, the MCS-7845 server has 4 virtual processors. So on the server where the main
router thread is running at full blast to do call processing, it is possible three other processors are near
idle. In this situation UC Manager can get into Code Yellow state even when total CPU usage is 25 to 30
percent. (Similarly MCS-7835 server with two virtual processors, UC Manager could get into Code
Yellow state at around 50 to 60 percent of CPU usage.
Use the following to query perfmon counters:
• SOAP APIs:
–
Perfmon counters
– Device information
–
DB access
– CDR access
•
SNMP:
– CISCO-CCM-MIB—ccmPhoneTable, ccmGatewayTable, etc.
–
Download historical information by using RTMT Trace/Log Central.
– Cisco AMC Service PerfMonLog is enabled by default. This was deprecated in Cisco Unified
CM Release 6.0 because Cisco RIS Data Collector PerfMonLog was introduced.
–
Cisco RIS Data Collector PerfMonLog was disabled by default in Cisco Unified CM Release
5.x and enabled by default in Cisco Unified CM Release 6.0.
CodeYellow
CodeYellow state occurs when the ccm process is so overloaded that it cannot process incoming calls
anymore. In this case, Cisco Unified CM initiates call throttling. This does not mean that one processor
CPU usage is at 100 percent and the remaining processors are operating at 0 percent in RTMT.
Since the main thread can run on processor A for 1/10th of a second and processor B on the next 2/10th
of a second, etc., the CPU usage shown in RTMT would be more balanced. By default RTMT shows
average CPU usage for a 30-second duration.
You can configure the CodeYellow alert so that once it occurs, the trace files can be downloaded for
troubleshooting purposes.
The AverageExpectedDelay counter represents the current average expected delay for handling any
incoming message. If the value is above the value specified in “Code Yellow Entry Latency” service
parameter, CodeYellow alarm is generated. This counter is one of key indicator of call processing
performance issue.
If you see CodeYellow, but the total CPU usage is only 25 percent, it is because Cisco Unified CM needs
one processor for call processing. When no processor resource is available, CodeYellow may occur even
when the total CPU usage is only around 25 to 30 percent in a 4-virtual processor server. Similarly on a
2 processor server, CodeYellow is possible around 50 percent of total CPU usage.