CloudWatch Monitoring free disk space on a tempdb for EC2 instance using Instance Store partitioned as ephemeral drives


What metrics or information is necessary to collect in order for the CloudWatch Agent to publish metrics to CloudWatch for an EBS volume partitioned into two Instance Stores for tempdb's with SQL Server?

已提问 10 天前133 查看次数
1 回答

Hi Andrew,

It appears there was a mix-up between EBS volumes and Instance Stores, which are actually distinct types of AWS storage as you might already know. Please let me know if I misunderstood it.

Coming back to the question again, The CloudWatch Agent can collect the metrics like Disk Reads/sec, Disk Writes/sec, Disk Read Bytes/sec, Disk Write Bytes/sec, % Disk Time, and metrics related to free disk space such as "Free Megabytes" or "% Free Space" on Windows servers. These metrics are standard for Windows Performance Monitor and can be configured for collection by the CloudWatch Agent through its configuration file. I think following metrics are important in order to monitor the disc for tempdb:

  • Disk Reads/sec and Disk Writes/sec: To gauge the I/O throughput, important for understanding how frequently data is being read from or written to the tempdb.
  • Disk Read Bytes/sec and Disk Write Bytes/sec: Measure the volume of data transferred, which helps in assessing the workload handled by the tempdb.
  • % Disk Time: Indicates the percentage of time the disk is active, providing insight into disk utilization.
  • Current Disk Queue Length: Shows the number of pending disk operations, useful for identifying bottlenecks.
  • Free Megabytes or % Free Space: Essential for monitoring available disk space to ensure tempdb has sufficient room to operate efficiently.

Hope this helps.

已回答 10 天前
  • Found this link on re:post to configure the CW Agent on Windows:

  • Hi Awaneedra, Thank you for the follow-up response. Essentially, this may provide a better overview for let's say a server has volumes attached for drives A-Z on a r5.4xlarge instance running MSSQL Server. Since the instance comes 2 NVMe drives, one of these SSD drives is used to host TempDB and has two partitions for "A" Drive. Because of this, however, CloudWatch does not pull in the metric for %FreeSpace or %LogicalDiskSpace like it currently does with the agent for drives B-Z. I am interested if taking the drive name like /dev/sda2 or xvdc is needed to be listed in resources line specifically in the CloudWatch agent to make monitoring the two partitions available and create alarms in CW for low storage space?

    Script and more info

您未登录。 登录 发布回答。