Docs

Elastic Compute Service

Instance diagnosis

Diagnosis item overview

This topic summarizes diagnosis items, detection items, repair suggestions, and other services provided by the instance diagnosis feature.

Computing service diagnosis

Diagnosis metric ID	Diagnosis item description	Impact	Repair suggestion
Instance.UtilizationHigh.CPU	Check whether the average CPU usage, single CPU usage, or single process usage is too high.	High CPU usage or load may cause slower system response, instance lag, and no response from programs and other issues.	Troubleshooting and solution for high CPU usage or load of ECS instance with Linux system
Instance.HighUtilization.Memory	Check whether the memory usage rate of the instance is too high.	Too high memory usage may cause: The system swaps memory frequently. This can further lead to slower program response and obvious performance degradation. The memory is exhausted. You can not allocate sufficient memory to new processes or tasks, which causes no response from the program and system crashes.	Troubleshooting and solution for high memory usage of ECS instance with Linux system
Instance.SystemLockUp	Check whether the Soft Lockup exception occurs.	If the Soft Lockup exception occurs, the system may stop all responses (unresponsive or frozen), and fail to process any input or output and does not allow logins.	Solution for the Softlock up issue of virtual machines
Instance.SystemLockUp		Check whether the Hard Lockup exception occurs.	If the Hard Lockup exception occurs, the system may stop all responses (unresponsive or frozen) and fail to process any input or output.	Submit a ticket
Instance.LoginAuth	Check whether the login of the instance has permission.	Because an incorrect password/key is entered, the login to instance fails.	Resetting login password
Instance.HostCheck	Check whether exceptions occurred on the host machine (host) of the instance.	Exceptions occurred on the host of the instance. This may cause risks for instance operations.	Resolution and handling of operation and maintenance events
Instance.RecentUtilHigh	Check whether the IOPS load reaches 85% of the upper limit.	When the IOPS load in the instance reaches 85% of its limit, this can affect the overall response time and efficiency of the instance.	Troubleshooting and solution for high disk BPS/IOPS load of ECS instance with Linux system
		Check whether the BPS load reaches 85% of the upper limit.	If the BPS of the instance reaches 85% of the upper limit, the speed of data transmission is affected.	Troubleshooting and solution for high disk BPS/IOPS load of ECS instance with Linux system
		Check whether the instance CPU load and usage is too high.	If the CPU load of the instance is too high, computing resources may be insufficient or too many zombie processes exist. This can further cause the instance to lag.	Troubleshooting and solution for high CPU usage or load of ECS instance with Linux system
Instance.DependenceCheck	Check the connectivity of key components that the instance depends on, and whether these components are functioning normally.	If communication exceptions between instance components and their dependent components occur, an in-band crash event is triggered. This can further trigger the OOM event for the instance memory.	Submit a ticket
Instance.CPUSplitLock	Check whether the instance frequently triggers Split Lock.	If an instance frequently triggers Split Lock, issues may occur, such as too high CPU usage of the instance and program lag.	Solution for Windows/Linux system performance degradation caused by frequent split lock triggers in the code

Instance configuration management diagnosis

Diagnosis metric ID	Diagnosis item description	Impact	Repair suggestion
Instance.BootFailure	Check whether the file system corrupts.	If the file system corrupts, the instance may fail to start up or be logged in.	Solution for handling file system corruption of Linux instances
		Check whether the Virtio version of the third-generation ECS instance is too low.	If the Virtio version is too low, the instance may fail to start up.	Solution for fixing third-generation instances that fail to start due to Virtio driver issues
		Check devices in fstab for exceptions.	If exceptions of the device mounting in fstab occur, the instance may fail to start.	Solution of fstab configuration repair for Linux instances
Instance.CreateFailure	Check instance creation exceptions due to insufficient quotas or resource inventory.	Insufficient resource quota can cause instance creation exceptions.	View and adjust instance resource quotas
Instance.CreateFailure				Insufficient resource inventory will cause instance creation exception.	Please try again later, or select another instance type to create.

Storage service diagnosis

Diagnosis metric ID	Diagnosis item description	Impact	Repair suggestion
Instance.HighUtilization.Disk	Check whether the capacity usage and inode usage of the disk mounted to the instance are too high.	If the usage is too high, this may cause the applications in the instance to fail to create new directories or files.	Troubleshooting and solution for high disk load of ECS instance with Linux system Troubleshooting and solution for high disk load of ECS instance with Windows system
Instance.IOHang	Check whether the system volume of the instance is experiencing IO latency or IOHang.	For volumes with increased I/O latency or IOHang, your read and write operations may fail.	To prevent this issue from occurring again, reduce the read/write frequency of the volume or upgrade the current volume to a higher-performance cloud volume type. For the metrics of volume read/write performance, refer to Block storage specifications.

Network service diagnosis

Diagnosis metric ID	Diagnosis item description	Impact	Repair suggestion
Instance.NetworkPort	Check whether the security group of the instance allows key ports to be accessible.	If you do not allow common ports for the security group during the instance runtime, you may fail to connect the instance normally.	Allow common key ports in the network security group of the instance
Instance.NetworkPort		Check whether the key rules of the instance Network ACL allow access.	If you do not allow common ports for the Network Access Control List during the instance runtime, you may fail to connect the instance normally.	Allow key ports in the Network Access Control List (ACL) for the instance
Instance.DDoSStatus	Check whether the instance EIP is under a DDoS attack.	If the instance EIP is under a DDoS attack, the instance is inaccessible.	Purchase DDoS protection products to defend against DDoS attacks
Instance.NetworkReachable	Check the ICMP connectivity of instances in the same VPC.	If ICMP connectivity exceptions occur to an instance, the instance may fail to connect the network.	Allow the inbound and outbound ports for the ICMP protocol in the instance network security group Allow the inbound and outbound ports for the Network Access Control List in the instance network security group
Instance.ARPing	Check whether the instance does not respond to ARPing requests.	If the instance does not respond to ARPing requests (failed response), this may affect instance network communication and thus the normal usage of the instance.	Restart the instance or check whether the instance network configuration is correctly configured.
Instance.HighUtilization.Network	Check whether the instance reaches upper limit of the network PPS.	When an instance network PPS (Packets Per Second, or packets per second) reaches the upper limit, this indicates the total internal and external network traffic of the instance exceeds the upper limit of the base network bandwidth for the instance type. This can cause issues for businesses, such as service disruption and exceptions. tip To obtain the data of current metric, refer to Viewing basic/operating system monitoring data.	Change to an instance type with higher network bandwidth capability
Instance.HighUtilization.Network		Check whether the instance reaches upper limit of the network BPS.		When an instance network Bits Per Second (BPS) reaches the upper limit, this indicates the total internal and external network traffic of the instance exceeds the upper limit of the base network bandwidth for the instance type. This can cause issues for businesses, such as service disruption and exceptions. tip To obtain the data of current metric, refer to Viewing basic/operating system monitoring data.
Instance.HighUtilization.Session.Max	Check whether the instance reaches upper limit of the session.	When an instance reaches the upper limit for its network session, it may fail to establish network card sessions properly or exceed the session count limit. This can further affect the instance network connectivity or throughput. For example, remote connections to the instance may fail or the instance network speeds may become much slower. tip To obtain the data of current metric, refer to Viewing basic/operating system monitoring data.	Restart the instance to clear the current network session state and release session resources that exceed the limit. If you can not resolve the issue by restarting the instance, we recommend that you further optimize the network session configuration. For example, you can adjust the maximum connections of the connection pool to avoid too many concurrent connections. If you can not resolve the issue with above solutions, we recommend that you change to an instance type with higher network bandwidth capability. By doing this, you can improve network bandwidth and the capability to handle sessions for the instance.

Operation audit diagnosis

Diagnosis metric ID	Diagnosis item description	Impact	Repair suggestion
Instance.UnexpectedFee	The operational behavior audit that involves instance billing to check whether billing method of the instance is changed during the specified time.	If the billing method of the instance is changed during the specified time, you need to confirm whether this is scheduled to avoid effects on businesses.	View event records
Instance.UnexpectedStatus	Check changes of instance status caused by unexpected operations, such as unexpected starting or stopping of the instance.	If the instance state changes due to unexpected operation events during the specified time, you need to confirm whether this is scheduled to avoid effects on businesses.	Querying and handling events
Instance.UnexpectedSG	Check whether the associated security group or security group rules of the instance are changed.	If you add an instance to a security group within the specified time, security risks of the instance may occur. If you remove an instance from a security group within the specified time, you may fail to access the instance.	View event records
Instance.UnexpectedSG			View event records	If you change a security group rule within the specified time, you may fail to access the instance.

Security risk diagnosis

Diagnosis metric ID	Diagnosis item description	Impact	Repair suggestion
Instance.SecurityRisk	Check whether security risks of the instance exist.	If the instance has security risks, you may fail to use the instance normally.	To view security status Enable security enhancement

Fee management diagnosis

Diagnosis metric ID	Diagnosis item description	Impact	Repair suggestion
Instance.Fee	Check whether the subscription-based instance has expired.	When a subscription instance expires, it is shut down and its services stopped. You can not access this instance afterwards.	Instructions on resource status after subscription instances expire Instructions on renewal and unsubscription
		Check whether the pay-as-you-go instance stops due to arrears.	If your account is in arrears, your instance may be stopped and its services stopped. You can not access this instance afterwards.	Instructions on resource status of pay-as-you-go instances during arrears Bill settlement
		Check whether the components of the instance are overdue.	If your account is in arrears, services of components associated with the instance are stopped, which can cause effects on the instance access afterwards.	Bill settlement

Diagnosis of related configurations in the instance operating system (Linux)

Diagnosis metric ID	Diagnosis item description	Impact	Repair suggestion
GuestOS.CPUUtil	Check whether the instance CPU usage is too high.	High CPU usage may cause slower system response, instance lag, and no response from programs and other issues.	Troubleshooting and solution for high CPU usage or load of ECS instance with Linux system
GuestOS.MemUtil	Check whether the memory usage rate of the instance is too high.	Too high memory usage may cause: The system swaps memory frequently. This can further lead to slower program response and obvious performance degradation. The memory is exhausted. You can not allocate sufficient memory to new processes or tasks, which causes no response from the program and system crashes.	Troubleshooting and solution for high memory usage of ECS instance with Linux system
GuestOS.OSOOM	Check whether the instance system has an OOM event.	If an OOM event occurs within the instance, you may fail to allocate the memory.	OOM handling methods for Linux instances
GuestOS.DiskUtil	Check the disk-related configurations and usage of the instance for exceptions.	If the version of the instance operating system is too low (exception), this may cause IOHang.	How to avoid IOHang issues caused by an outdated kernel version of the operating system
GuestOS.FileSystems	Check whether the device in fstab exists.	If a device configured in the /etc/fstab file of the instance does not exist, the instance may fail to start.	Solution of fstab configuration repair for Linux instances
		Check whether disks in fstab are correctly mounted.		If the disk is not mounted in /etc/fstab, the instance may fail to start.
		Check whether the format is valid.		If the file format of /etc/fstab is invalid, the instance may fail to start.
GuestOS.SSHServiceStatus	Check whether you enable the multi-factor authentication in the configuration of SSH access permission.	If you enable multi-factor authentication in the configuration of SSH access permission, both the password and key are required when you log in to the instance. Otherwise, your login via ECS Terminal may fail.	Solution for login failure due to SSH configuration
		Check whether SSH service of the instance allows root account login.		The instance SSH service currently does not allow login by the root account. This may cause that you can not access the instance via SSH with the root account.
		Check whether the key files or directories required for SSH access exist.		If the key file or the directory for the instance SSH service is missing, you may fail to access the instance via SSH.
		Check whether the key file format required for SSH access is abnormal.		If the format of the key file for the instance SSH service is invalid, you may fail to access the instance via SSH.
		Check the access permission of key files or directories for SSH for exceptions.		Improper configuration of file access permissions required by the instance's SSH service may result in the instance being inaccessible via SSH.
		Check whether the SSH listening port is configured consistently.		The address and port currently being listened to by the instance SSHD process are inconsistent with the configured address and port, which may cause SSH to fail to connect to the expected address and port.
		Check whether the OpenSSL version matches.	If the OpenSSL version of the instance and the SSH service do not match, this may cause the instance inaccessible via SSH.	How to resolve the issue of being unable to remotely log in to the instance via SSH after upgrading OpenSSL
GuestOS.DriverStatus	Check whether the version of Virtio in the instance system is correct.	If the Virtio driver version of the instance system is low (exception), issues may occur. For example, the instance may fail to run normally or the instance disk may fail to expand online.	Solution for fixing third-generation instances that fail to start due to Virtio driver issues
GuestOS.CloudInitService	Check the status of the Cloud Init.	If the Cloud Init driver of the instance is in an exception status, this may cause failure of related system configurations during instance system initialization and inaccessible instance.	Installing Cloud Init in the Linux system
GuestOS.CloudInitService		Check the status of the Metaserver.	If the metadata service (Metaserver) of the instance fails to connect properly or times out, the instance metadata may be inaccessible.	Recommended network configuration solutions
GuestOS.SystemConfig	Check whether SELinux is enabled	If the SELinux service on the instance is currently enabled, you may fail to log in to the instance via SSH remote connection.	Recommended network configuration solutions
		Check whether the limits settings are correct.	If some configurations in the instance system file /etc/security/limits.conf exceed the system default values, you may fail to log in to the instance via remote connection.	Methods to change the system limit parameters for Linux instances
		Check whether the large-page memory of an instance is configured to be too large.	If the large-page memory of an instance is configured to be too large, this may cause insufficient memory of the instance and failure to log in remotely.	Methods to adjust large-page memory for Linux instances
		Check whether the instance's kernel configuration for non-maskable interrupts is inappropriate.	Improper configuration of Non-Maskable Interrupt (NMI) in the instance kernel will cause unexpected kernel panic and instance restart.	How to adjust kernel argument kernel.unknown_nmi_panic for a Linux instance
		Check whether the number of processes running on the instance exceeds two-thirds of the maximum process count.	The number of processes currently running in the instance system exceeds two-thirds of the maximum number of processes (kernel.pid_max), which may lead to the inability to create new processes subsequently.	Linux instance starting the service prompts: task: Cannot allocate memory
		Check whether the NVMe disk IO timeout configuration is reasonable.	The IO read/write timeout period for NVMe disks in the instance's system files is too short, which may lead to the disk becoming read-only due to IO timeout, thereby causing subsequent write operations to fail.	Solution to improper configuration of I/O timeout parameters for Linux instance NVMe disks
		Check whether tcp_sack is enabled in the network.	The instance network tcp_sack was not enabled, which may affect the network performance of the instance.	How to enable TCP SACK for Linux instances to improve network performance
		Check whether the kernel argument configuration in the NAT environment is correct.	The kernel arguments related to the instance's NAT environment have an exception, which may cause an inability to connect to the instance via SSH and anomalies when accessing HTTP services on the instance.	Solution to the Linux instance kernel configuration issue causing abnormal NAT environment access to the instance
GuestOS.SystemUserPwd	Check whether the password of the key system account exists.	If the system account of the instance does not exist, you may fail to log in to the instance or start key processes.	Solution for missing system accounts in Linux instances
GuestOS.Firewall	Check the system firewall status of the instance.	If you enable the firewall for the instance and set rules to block external access, you may fail to access the instance via the remote connection.	Enabling or disabling system firewall in Linux instances
GuestOS.NetworkStatus	Check the instance in-band access to the OpenAPI gateway for exceptions.	If an exception of in-band access to the OpenAPI gateway occurs, this may cause other exceptions, such as the loss of in-band monitoring and inaccessibility to OpenAPI.	Recommended network configuration solutions
		Check whether related processes exist if the network is configured with DHCP.	If the DHCP service process of the instance network interface is disabled or does not exist, the instance may fail to renew its IP address after the lease expires, which can cause network interruption.	Solution for missing networks on Linux systems
		Check whether network-related ports 80 and 443 are in a listening status.	The network-related ports of the instance may be in a listening status.	We recommend that you make sure the process listening on the port is scheduled to prevent malicious processes from listening.
		Check whether the key service sshd is in a listening state.	If the sshd service on the instance is currently disabled, you may fail to log in to the instance via remote connection.	Solution for SSH start failure
		Check whether you enable the maximum queues per network interface.	If the maximum queues per network interface are not enabled for the instance, this may affect instance network performance.	Recommended network configuration solutions
		Check whether the IPv4 address of the instance or the corresponding subnet mask configuration is correct.	The IPv4 address or corresponding subnet mask configuration of the instance is incorrect, which may prevent the instance from connecting and being accessed properly.	Linux instance configuration method for setting a static IP address (Static mode)
GuestOS.TimeSyncService	Check the status of the instance's time synchronization service.	The time synchronization service of the instance is not working properly or is misconfigured, which may cause the time within the instance system to deviate from the actual standard time, thereby impacting the normal operation of some applications within the instance.	Time synchronization service overview

Diagnosis of related configurations in the instance operating system (Windows)

Diagnosis metric ID	Diagnosis item description	Impact	Repair suggestion
GuestOS.WinCPUUtil	Check whether the instance CPU usage is too high.	High CPU usage may cause slower system response, instance lag, and no response from programs and other issues.	Troubleshooting and solution for high CPU usage or load of ECS instance with Windows system
GuestOS.WinMemoryUtil	Check the disk-related configurations and usage of the instance for exceptions.	Too high memory usage may cause: The system swaps memory frequently. This can further lead to slower program response and obvious performance degradation. The memory is exhausted. You can not allocate sufficient memory to new processes or tasks, which causes no response from the program and system crashes.	Troubleshooting and solutions for high memory usage in ECS instances running on Windows systems
GuestOS.WinMemoryUtil				When the instance reserves too much memory for hardware, this may cause increased memory load and instance lag.	Solution for Windows instance lag due to excessive hardware-reserved memory
GuestOS.WinDiskUtil	Check whether the usage rate of the system volume capacity is too high.	If the usable space of the system volume (drive C) in the instance is insufficient, this may cause the system to run slowly or the instance fail to start.	Expanding volumes for Windows instances
		Check whether the system volume is in an exception status.	If the instance system volume (drive C) is in an exception status, issues may occur, such as failed restart of the instance and failed driver installation for the instance.	Solution for system volume exceptions of Windows instances Rolling back the volume with the snapshot feature tip You can also execute the chkdsk command to check the status of all volumes, identify, and locate volumes with exceptions. Then, execute the sfc command for manual repair or replacement.
		Check whether the version of the Virtio driver is too low.	If this instance has a duplicate disk uniqueID due to a low virtIO driver version, this may cause loss of disk data on the instance when you reset the disk.	Methods to update Virtio driver in Windows instances
GuestOS.WinSystemConfig	Check whether key system files exist.	If key system files in the instance system directory (C:\\Windows\\) are missing, this may cause black screens or operation exceptions after instance login.	Solution for remote login failure for ECS instance with Windows system due to black screen
		Check whether the system is in activation status.	If you do not activate the Windows operating system for the instance, some Windows services may not function normally.	Methods to activate the Windows operating system for ECS instances
		Check whether the system patches are applied correctly.	If you install inappropriate system patches for the instance, abnormal system restarts or crashes may occur.	Uninstalling system patches for Windows instances
		Check whether the software protection platform service runs normally or is started.	If you fail to start or run the Software Protection Platform Service (SPPSVC) normally for an instance, you may not activate the Windows system and access activation settings.	Methods to activate the Windows operating system for ECS instances
		Check whether the activation server of the key management service is correctly connected.	If the instance can not connect to the activation server for Key Management Service (KMS), this may cause failed activation of the instance.	Solution for activation failure of Windows-based ECS instance
		Check whether the key management service matches the Windows version.	If you activate the Windows operating system for an instance with the Key Management Service (KMS), the activation may fail if the activation key used by the KMS client does not match the Windows version.
		Check the status of the crash dump configuration.	When the crash dump function of the instance is disabled and the system restarts due to exceptions or blue screen occurs, the system can not save the related information for troubleshooting.	Solutions for Windows system unexpected restarts and blue screens
		Check whether the version of the operating system is too low.	If the operating system version of the instance is too old, Volcano Engine and Microsoft no longer maintain it, which may affect your normal use. It is recommended to upgrade the operating system.	Change the operating system
GuestOS.WinDriverStatus	Check whether driver residues exist.	If residual files in the disk filter driver of the instance exist, this may prevent the instance from recognizing newly mounted disks.	The method for checking residual items in the disk driver of an ECS instance with Windows system
GuestOS.WinDriverStatus		Check whether the version of the Virtio driver is too low.	If the Virtio driver version of the instance is too low, this may cause risks, such as blue screens, network packet loss, or disk data loss.	Methods to update Virtio driver in Windows instances
GuestOS.WinSystemProcess	Check the status of key system processes (RDP status).	If the RDP service of the instance is disabled or corrupted, the instance may be inaccessible via RDP.	We recommend that you restart or reinstall the remote desktop connection (RDP). If you need to reinstall RDP, execute the commands: tip Execute the following commands in cmd or PowerShell. net stop TermService net start TermService
		Check the status of key system processes (firewall of RDP).	If the remote desktop connection service (RDP) for the instance is blocked by the firewall, this may cause the instance to be inaccessible via RDP.	Configuring firewall policies for ECS instances on Windows Server
		Check the status of key system processes (RDP license).	If the license for multi-user remote desktop service of the instance has expired, you may fail to use the RDP normally and to access the instance via remote desktop.	Solution for remote login failure for ECS instance with Windows system due to expired service authentication
		Check the status of key system processes (the metadata service of the instance).	If the metadata service (Metaserver) of the instance currently fails to connect properly or times out, the instance metadata may be inaccessible.	Instance metadata overview tip Check whether you configure the instance firewall to block the IP address 100.96.0.96. If this IP address is blocked, allow the IP address in the firewall configuration before you access the metadata service. The method for configuring the firewall policy of an ECS instance with Windows Server system
GuestOS.WinSystemInit	Check the initialization status of the system.	If you restart the instance too early when creating it, this may discontinue the initialization process of the system preparation service (SysPrep). You may fail to start the instance due to some key configurations of the operating system uncompleted.	Change the operating system
GuestOS.WinSystemUser	Check whether the system account and password exist.	If the Administrator account of the instance does not exist, you may fail to log in to the instance or start key processes.	Solution for the nonexistence of system users in Windows system instance
GuestOS.WinNetworkStatus	Check the port status of the system.	If ports of the instance system are not allowed or the firewall is enabled, this may cause the RDP unavailable to access the instance.	Methods to start RDP in Windows operating system instance
		Check the status of the port for the remote desktop protocol service.	If the port used by the instance Remote Desktop Protocol (RDP) is occupied by other processes, a port conflict then occurs. You may fail to access the instance via remote desktop.	Solution for port conflict during remote login to a Windows system instance
		Check the status of the instance network interface.	If the network interface of the instance is currently unavailable, you may fail to access the instance via the remote connection.	Solution for unavailability of the system network interface in a Windows instance
		Check whether related processes exist if the network is configured with DHCP.	If the process of DHCP service of the instance network interface is in disabled status, you may fail to rent the IP address for the instance again after the IP address expires. This can discontinue the instance network.	Methods to install and configure DHCP servers for Windows instances
		Check whether the network interface is assigned with an IP address.	If the network interface of an instance fails to obtain an IPv4 address, this may cause the service to be inaccessible.
		Check the status of the network proxy.	If the instance is configured with a network proxy, this may cause inaccessible service.	Methods to reset the agent for Windows instances
GuestOS.WinLogin	Check the login status.	Because you enter incorrect passwords multiple times, the user account is locked and you can not log in to the instance with the account.	Methods to adjust the threshold for the account lockout policy in Windows instances
GuestOS.WinThirdPartSoftware	Detect the status of installed third-party software.	If you install a third-party antivirus software for the instance, some management operations (such as password reset and remote connection) may fail and thereby cause instance exceptions.	Methods to uninstall or delete applications and programs in Windows instances

Diagnosis of heterogeneous computing GPU

Diagnosis metric ID	Diagnosis item description	Impact	Repair suggestion
GuestOS.RdmaAgentService	Check the status of the RDMA network configurator.	If the rdma-agent.service of the instance is in an exception status, issues may occur. For example, the RDMA network is unavailable, or RDMA network performance falls short of expectations.	Solution for exceptions of RDMA network configurator
GuestOS.Nvidia.DriverStatus	Check the version of the GPU driver.	If the GPU driver is not detected and does not respond, you may fail to use GPU normally.	Installing GPU drivers
GuestOS.Nvidia.FabricmanagerService	Check the status of the nvidia-fabricmanager.service on the instance.	If the nvidia-fabricmanager.service of the instance is in an exception status, the GPU can not function normally.	Solution for nvidia-fabricmanager.service status exception
GuestOS.Nvidia.ModStatus	Check whether the kernel modules required for instance GPU are loaded.	If kernel modules required for GPUs of the instance are not loaded, partial GPU features may be unavailable or the GPU performance may be degraded.	Solution for unloaded kernel modules required by NVIDIA
GuestOS.Nvidia.NicConfig	Check whether the RDMA network interface in the instance is normally configured.	If the on-chip configuration of instance RDMA network interfaces is incorrect, the RDMA network interface may fail to function correctly or the performance of RDMA network interface may be degraded.	Submit a ticket
GuestOS.Nvidia.RailOptimized	Check whether the RDMA network interface in the instance is configured for multi-track networking.	If the RDMA network interface in the instance is not configured for multi-track networking, the RDMA network performance may be degraded.	Submit a ticket
GuestOS.MonitorStatus	Check whether Volcano Engine monitoring services are operating normally within the instance.	If the monitoring service in the instance is not installed or is in an exception status, this may cause delayed detection of instance exceptions.	Installing plugins
GuestOS.Nvidia.NvmlStatus	Check the NVIDIA NVML library in the instance for exceptions.	If the NVIDIA NVML library within the instance fails to link correctly, this may cause delayed detection of GPU exceptions in the instance.	Solution for link exceptions in NVML link libraries
GuestOS.Nvidia.OfedStatus	Check whether the driver for the RDMA network interface is correctly installed within the instance and whether the correct driver version of RDMA network interface can be obtained.	If the driver for RDMA network interface in the instance is not detected, you may fail to use RDMA network interface normally.	Submit a ticket
GuestOS.Nvidia.Gdr	Check the GDR performance of the GPU for the instance.	If an exception occurs, this may cause low GDR performance between the RDMA network interface and the GPU of the instance. If the performance test fails, this may affect the GDR connection between the RDMA network interface and the GPU of the instance.	Submit a ticket
GuestOS.Nvidia.RdmaNic	Check the performance and connectivity of RDMA network interface.	The performance of the RDMA network interface is low or the RDMA network interfaces within the instance can not connect.	Submit a ticket
GuestOS.Nvidia.NcclTest	Check whether the performance of NVIDIA nccl-tests on one instance is abnormal.	The performance of the instance in NVIDIA nccl-tests results is low.	Submit a ticket
GuestOS.Nvidia.H2dBandwidthTest	Check the bandwidth performance between GPU and CPU for exceptions.	The bandwidth performance between the GPU and the CPU is low.	Submit a ticket
GuestOS.Nvidia.P2PBandwidthTest	Check the bandwidth performance and connectivity between GPUs for exceptions.	The bandwidth performance between GPUs in the instance is low or the bandwidth connection fails.	Submit a ticket
GuestOS.Nvidia.DcgmTest	Check whether exceptions occur in the NVIDIA DCGM.	If exceptions occur, this may cause other NVIDIA GPU exceptions.	Submit a ticket
GuestOS.Hpc.ConnectivityTest	Check the connectivity of RDMA network interfaces and switches within the cluster, and the connectivity of the switches.	RDMA network interfaces in the instance are not connected to the cluster.	Submit a ticket
GuestOS.Hpc.HpcEnvironmentTest	Check whether HPC-optimized instances are in the same HPC cluster.	If the selected instances are not in the same HPC cluster as other instances, resource usage may be insufficient or uneven.	Submit a ticket
GuestOS.Hpc.NcclTests	Check the connection status of instances within the cluster in nccl-tests.	If the connection of NVIDIA nccl-tests within the cluster fails, conduct troubleshooting and analysis based on the diagnosis results of the performance exceptions.	Submit a ticket
GuestOS.Hpc.NcclTests		Check the nccl-tests performance between instances within the cluster.	If the performance shown in the NVIDIA nccl-tests within the cluster is low, conduct troubleshooting and analysis based on the detection results of the exception diagnosis items of the instance performance. If there are other loads on the GPU during the test, ignore this error. If other GPU issues exist, fix and try again.	Submit a ticket

Last updated: 2026.10.14 17:46:43