Response to facility trouble

Are storage devices redundant?/What operation is performed in the event of storage device trouble?

The following two types of configuration are used for storage devices providing the disk service of SAKURA Cloud.
The provided configuration differs depending on factors such as the timing of service implementation, etc.

● Configuration of storage pool type

In this configuration, one set consists of a controller and a storage pool that is an aggregation of disks.

Normally, service is continued using the controller at the active side. If a controller error is detected, it immediately switches to the controller at the standby side.
Also, disks in the storage pool have redundancy using RAID. Therefore, continuation of service is not affected even when several disks have trouble at the same time (normally, malfunctioning disks are replaced within 24 hours).

● Configuration of the active standby type

In this configuration, active standby is performed for one set of two devices with exactly the same settings.

Normally, service is continued via devices on the active side. If an error is detected, the configuration automatically switches to the standby side.

*Normally, several dozen seconds to several minutes is required for the switching operation. Disk I/O is delayed during this period (the disk is not disconnected).

*Figures are abbreviated in order to facilitate explanation. Therefore, some portions differ from actual conditions.

Are archives created using the [Archive] function affected by trouble of the [Disk] function?

The storage machine storing archives and the storage machine storing disks exist separately. Therefore, the scope of impact at the time of trouble at storage machine is limited to the provided functions.

Is redundancy used for the host server that stores the virtual server?/What kind of operation is performed if the host server has trouble?

All virtual disk images connected to the virtual server which is operated by the host server are stored in storage. The images are read/written via the storage network. Therefore, the host server does not hold data other than memory information.

If trouble occurs at the host server and the host server stops, the stored virtual servers also stop at the same time. Via the automatic recovery program at the host server management server, an attempt is made to restart the previously operating virtual servers at other host servers. Since this is separate from the virtual disk, data in the disk is not cleared by the host server. (There is the possibility of logical data corruption due to sudden OS stoppage).

*Figures are abbreviated in order to facilitate explanation. Therefore, some portions differ from actual conditions.

Is the network redundant?

Duplication is used for all lines, network devices, interfaces, etc. at important channels; for example, the storage network and networks in which virtual servers communicate with external destinations.

If trouble occurs in the duplicate portion, there is instantaneous switchover to the channels and devices on one side. This minimizes the impact of services.

How is notification given in the event of trouble?

If trouble occurs on SAKURA Cloud, trouble information is immediately listed.

Please refer to Support information website.

Also, when trouble occurs at the host server that stores the customer virtual server, an email with the following subject is sent to notify customers of the trouble information.

 Notice: Trouble Detected

If monitoring of the host server fails, an email notification is immediately sent to all related users.

 Notice: Error Detected

An email notification is sent to all related users at the start of recovery for resources such as the switch and virtual server stored in the host server where trouble occurred. Also, trouble information is listed on Support information website.

 Notice: Recovery from Error

When recovery for resources such as the switch and virtual server is finished, an email notification is sent to all related users. Also, the details of trouble information listed at the time of occurrence are updated.