Use Spotlight on SQL Server Enterprise to monitor performance health of Unix/Linux servers that host your SQL Server databases. Spotlight offers:
Spotlight can monitor the following operating systems:
The required connection parameters are: Server Address, User, Password and Port Number.
Our Release Notes document outlines the required user permissions in order to monitor Unix/Linux servers in Spotlight.
Once connected, Spotlight Overview page offers 6 panels of categorical monitoring components:
Details of each monitoring component category:
Component and Flow
Network Users - The number of users logged on remotely to this Unix/Linux machine
a. Established - The total number of TCP/IP connections to this machine in the ESTABLISHED state.
c. Close_Wait - The total number of TCP/IP connections to this machine that are in the CLOSE_WAIT state, where the remote end of the connection has shut down, and is waiting for the local end to do the same.
a. Pkts/S - The rate of network packets being received and sent for the network card that has the highest rate of packet transfers.
b. Errors/s - The rate of erroneous network packets being received and sent for the network card that has the highest rate of packet transfers.
OS - The flavor of Unix/Linux that is running on this machine.
Uptime - The time elapsed since the last time the machine was booted, measured in days and hours.
CPUs - The number of working processors in the machine
Speed – The speed of the CPU measured in megahertz.
Type - The type of CPU installed on this Unix/Linux host.
CPU Usage - The value is the sum of the usage by the system and users.
Users % - The percentage of time that CPUs are in the User state.
System % - The percentage of time that CPUs are in the System state.
Queue Length % - The percentage of time that CPUs are in the Wait state, which occurs when the machine is actively waiting for an I/O operation to complete. A high Wait level may indicate badly configured disks.
a. Total - Shows the total number of processes that are running on the Unix/Linux machine.
b. Zombies - Unix/Linux expects a parent process to acknowledge the termination of any child process. If it fails to do so, the terminated child process is classified by the kernel as a zombie. A high number of zombie processes indicates that one or more processes are not handling their child processes properly. A zombie’s parent process may need to be killed in order to eliminate its zombie child process.
c. Blocked - The number of processes waiting for some event or condition before they can continue execution.
a. Total RAM - Shows the amount of RAM in the Unix/Linux machine, present as RAM chips.
b. Free - The physical memory that is not in use on the machine.
c. Free Pct - The percentage of physical memory that is not in use on the machine.
d. Used - The gauge shows how much physical memory has been used.
a. Total Size - The total virtual memory of the Unix/Linux machine.
b. Free – The amount of virtual memory that has not yet been used.
c. Use Pct – How much virtual memory has been used, expressed as a percentage of total virtual memory
d. Used - Shows how much virtual memory has been used.
Total - The total swap space allocated by the operating system (in megabytes).
Used - The amount of swap space in use and the proportion of allocated swap space in use
Disks - Shows how busy the most active disk on the machine is as a percentage of maximum activity.
a. Total - The theoretical total size (in MB) of the fullest disk on the Unix/Linux host.
b. Used - The size (in MB) of the used space of the fullest disk on the Unix/Linux host.
Spotlight offers 20 monitoring alarms that can be acknowledged, snoozed and customized:
File System Space alarm
Indicates that the file system is full or filling up.
Memory - Physical Memory Available alarm
Raised when the available memory drops below a threshold.
Most Active Disk alarm
Triggered when the read/write rate of the most active disk exceeds a specified threshold.
Blocked Processes alarm
The Blocked Processes button in the CPU panel shows the number of processes waiting on information to become available in memory
CPU Busy alarm
The CPU Busy alarm occurs when the total CPU utilization of the system exceeds a threshold. The CPU may encounter a large number of requests, or you may have un-tuned SQL, which uses excessive amounts of CPU
Monitored Server - Unix/Linux Alarm Action Failure alarm
Raised when a (specified action) failed to evaluate
Monitored Server - Unix/Linux Alarm Evaluation Failure alarm
Spotlight failed to evaluate the specified alarm from the given collection.
Monitored Server - Unix/Linux Collection Execution Failure alarm
Raised when a (specified collection) failed to execute against the server.
Monitored Server - Unix/Linux Connection Failure alarm
Spotlight has failed to connect to the Unix/Linux Server.
Monitored Server - Unix/Linux Planned Outage alarm
Raised during a planned outage of the connection. Spotlight will resume monitoring the service at the end of the planned outage period.
Monitored Server - Unix/Linux Secondary Connection Failure alarm
Monitored Server - Unix/Linux Unsupported Version alarm
This version or flavor of Unix/Linux is not supported by Spotlight. Some collections may fail because Spotlight has not been tested against this Unix version or flavor yet. Use at your own risk.
Network Errors in alarm
Network Errors in represents the number of erroneous network packets coming in to the machine per second. The Network Errors in alarm may warrant investigation as even a low error rate can indicate network problems.
Network Errors out alarm
Network Errors out represents the number of erroneous network packets sent by the machine per second. The Network Errors out alarm may warrant investigation as even a low error rate can indicate network problems.
Paging in alarm
A large number of page ins may be a symptom of a large number of recent page outs.
Paging out alarm
If an active process asks the kernel for more memory than there is immediately available, the kernel will write old memory pages out to swap space. This is known as paging. To stop paging, make sure that there is enough RAM available to support the size of the processes you want to run
Swap Space alarm
If the total amount of swap space allocated to a Unix/Linux host becomes full, the machine may halt all processes, or critical actions may be prevented from occurring. Running out of swap space may indicate a runaway process or an under-configured machine.
Swapping in alarm
The Swapping In represents the number of processes swapped from disk per second. A machine that is swapping processes to or from disk is usually under-configured for its workload.
Swapping out alarm
The Swapping out represents the number of processes swapped to disk per second. A machine that is swapping processes to or from disk is usually under-configured for its workload.
Zombie Processes alarm
The Zombie button in the CPU panel shows the number of terminated child processes that have not been acknowledged by their parent process. A high level of zombie processes is indicative of the presence of an inferior program, or an entry in /etc/inittab (or equivalent) preventing init from completing its startup sequence.
Spotlight also includes partial support for monitoring SQL Server vNext hosted on Linux. See the Release Notes for more details.
Download a free trial edition of Spotlight from here.
Nice Blog, Gita!