Overview & Thematic Scope
This technical FAQ addresses the critical issue of high CPU utilization on Huawei MA5600T control boards, a common challenge for network engineers managing these multi-service access platforms. We provide definitive, expert answers to the most pressing questions about diagnosing and resolving CPU overload scenarios, helping you maintain service stability and performance.
Frequently Asked Questions
- Q1: What are the primary root causes of high CPU utilization on a Huawei MA5600T control board?
- The primary root causes are network-side or user-side packet attacks, routing loops causing protocol packet forwarding loops, and internal message flooding from service modules . Additionally, scenarios like mass automatic dial-up after a system reset, a large number of ONTs concurrently going online/offline, or malfunctions of the upper-layer switch can trigger CPU overload .
- Q2: What immediate symptoms indicate a CPU overload issue on the MA5600T?
- Symptoms include the system becoming unreachable, CLI commands returning a ‘System is busy’ error, and potential service board resets . In severe cases, the CPU usage can hit 100%, causing failure in managing the device via NMS and interruptions to communication between the control and service boards .
- Q3: What are the first diagnostic commands to run when investigating high CPU usage?
- Begin by using the DBWIN tool to diagnose the system. Enable DBWIN, then use ‘dbwin print 0x20 1’ to view the status of key system processes. If a specific PID is stuck in a waiting state (‘P’), you can use ‘dbwin send 0x20 3’ to identify it and ‘dbwin send 0x20 4 [PID]’ and ‘dbwin send 0x20 5 [PID]’ to release it and resolve the ‘System is busy’ error .
- Q4: How can I use DBWIN to resolve a ‘System is busy’ error caused by a locked process?
- First, enable DBWIN on the CLI using the command ‘dbwin enable’. Then, to get the problematic PID, execute ‘dbwin send 0x20 3’. The system will return a PID (e.g., 0xae). Release the locked PID by running ‘dbwin send 0x20 4 [PID]’ followed by ‘dbwin send 0x20 5 [PID]’. Finally, disable DBWIN with ‘dbwin disable’ to complete the process .
- Q5: Is there a permanent fix for CPU overload issues caused by packet flooding?
- Yes, Huawei has released patches for specific software versions to implement CPU overload control policies. These patches dynamically control packets based on CPU usage, ensuring critical tasks are prioritized. For example, versions V800R52SPC020 and later, as well as V800R62SPC118 and V800R007C00SPC307, include these protections . Upgrading to these or later versions is the recommended permanent solution.
- Q6: How do I collect essential system information for a thorough CPU utilization analysis?
- Collecting comprehensive data is crucial. Key steps include: 1) Using ‘debugging lswdrv bcm-cli BCM.0 show c’ to check LSW port packet reception/transmission; 2) Enabling DBWIN and using commands like ‘dbwin print 0x86 1’ (or 0x89 for older versions) to view packets sent to the CPU; 3) Checking service module messages with ‘dbwin print 0x65 1’ (or 0x95 for older versions); and 4) Exporting the configuration and database using ‘backup configuration’ and ‘backup data’ commands .
- Q7: What does the H802GPBD board reset issue have to do with CPU utilization?
- On MA5600T versions V8R6 and V8R7, H802GPBD boards may reset repeatedly due to a CPLD logic error triggered by intensive management message interaction (OMCI, PLOAM) during ONU discovery or frequent ONU online/offline events . This is often a symptom of a broader CPU management strain, and checking the CPLD version is key to identifying this specific cause.
- Q8: How can I prevent high CPU usage during mass ONT dial-up events?
- Implementing the official Huawei patches that control the rate of PPPoE and DHCP dial-up packets is the primary method . These patches ensure that during a mass automatic dial-up event (e.g., after system reboot), the CPU is not overwhelmed, allowing for service stability and system accessibility.
Leave a comment