Description

This chapter guides the administrator in handling SSD disk latency.

Precautions

1. The following situations do not support the automatic isolation of slow SSD:

• The system disk will not be automatically isolated when it is in high disk latency. Please contact Sangfor technical support for handling.

• Two nodes scenario does not support the automatic isolation of disks, so manual isolation is required, and disk data rebuilding is not supported after isolation. The hard disk must be replaced as soon as possible to ensure data integrity.

• The slow disk will not automatically be isolated when the node and hard disk are in maintenance mode.

• Before automatically isolating the hard disk, the virtual storage will check the integrity of the replica. If it detects that other nodes have copies in an abnormal state, isolating the current hard disk may cause the virtual machine on the hard disk to run in the None method. Manual isolation can be done according to the current situation, but this situation may cause business interruption.

• The latency disk that appears during the upgrade process will not be automatically isolated.

• The disk performing the bad sector repair task will not be automatically isolated.

• Currently, automatic isolation of the virtual witness node SSD is not supported.

2. The following situations do not support the permanent isolation of latency SSD disk:

• Under circumstances of 3 nodes and each node having 1 or 2 disk groups with three replicas, it doesn’t support permanent isolation of latency SSD disk.

• If the permanent isolation process is performed, it is found that the data cannot be completely reconstructed after the isolation process, and the latency SSD disk isolation process will not be executed.

• The disk of the physical witness node will not be handled with permanent isolation when the SSD disk is in latency, but temporarily re-attach the disk. When there is a latency SSD disk, must contact the customer to replace the disk as soon as possible.

3. The current version supports latency SSD disk handling but not slow SSD disk handling.
4. It only supports virtual storage scenarios. This function does not apply to non-virtual storage scenarios.
5. If there is a restart node operation, the slow disk processing service will take effect after half an hour.
6. Currently, it is not supported to identify the disk latency caused by the RAID card.
7. If multiple SSD disks are identified as latency, they will be processed one by one, and batch processing is not currently supported.
8. After the PCIe disk is kicked and the volume is deleted, the newly created volume cannot be scanned.
9. Multiple PCIe disks will affect each other. One latency disk may cause other disks to become slow.
10. Manually isolating SSD disks, the services of the entire disk group will be affected. Please operate with caution.

Prerequisites

None.

Steps

SSD disk latency handling is divided into three stages: temporarily isolated, permanently isolated, and manually isolated:

• When the SSD disk is flagged as a latency disk, you may proceed with temporary isolation (ensure it will not cause dual-point failure). The SSD disk will restore online after 5 minutes, then check again if the SSD disk is still detected as a latency disk.

• After being temporarily isolated three times, if it is still flagged to be a latency disk again, it will be permanently isolated (given that there is enough free capacity for data rebuilding).

• If the disk fails to isolate (for example, condition check fails), an Isolate Disk button will display on the page, and you can choose to isolate the latency disk forcibly.