JP4370818B2 - Power management method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a power management method for an uninterruptible power supply system configured by connecting a plurality of servers that receive power supply from the uninterruptible power supply on the same network as the uninterruptible power supply.
[0002]
[Prior art]
In general, an uninterruptible power supply (UPS) having a backup battery or a generator inside is always supplied from a commercial power supply to a system load such as a computer that requires uninterruptible power. It is widely used as a device that supplies power and allows the system to operate as it is even when a commercial power supply fails. The backup battery used in the UPS has a capacity that can operate the system for several minutes even if it has a small capacity, and the computer system can be safely shut down (terminated) during that time. Some large-capacity UPSs with an internal generator can operate the system for several days.
[0003]
FIG. 4 is a system configuration diagram showing an example of a conventional uninterruptible power supply system. In this system, a UPS 1 that can be connected to a network, and a plurality of server computers (hereinafter simply referred to as servers) 11 and 12 that receive power supply from the UPS 1 are connected by a hub (HUB) 2 on the same network such as a LAN. Constitutes an uninterruptible power system.
[0004]
FIG. 5 is a timing chart for explaining a conventional shutdown processing procedure. In the uninterruptible power supply system of FIG. 4, when an external commercial power supply fails or when the schedule is off, an OS shutdown command is sent from the UPS 1 to the servers 11 and 12 via the network, so that the OS shutdown in each server 11 and 12 is ensured. After being executed, it is necessary to stop the output of UPS1.
[0005]
However, in a system including the conventional UPS 1, regardless of whether or not the servers 11 and 12 are actually operating, all of the servers 11 and 12 set in the UPS 1 are shown in FIGS. The shutdown commands C1 and C2 as shown in FIG. 1 are sent from the UPS 1 via the network. As shown in FIGS. 5B and 5D, the output stop time set on the UPS 1 side (hereinafter, UPS output stop scheduled time) regardless of whether or not the OS shutdown processing in the servers 11 and 12 has ended. The power supply from the UPS 1 was stopped at the time t0 when e.g.) passed.
[0006]
On the other hand, on the server 11 and 12 side, like the server 12 shown in FIG. 5D, the OS shutdown process may be delayed due to some reason for the OS shutdown process for some reason. In such a case, if the power supply from the UPS 1 is not continued until time t1 or later, the power is cut off without the OS being shut down, and an abnormal termination process (dirty shutdown) may occur.
[0007]
Therefore, in a computer system in which a plurality of servers are connected in parallel as a load of one UPS 1, a shift occurs in information saved at the time of shutdown in each server, and as described in Patent Document 1 described later. One load is a master server, and a shutdown command needs to be transmitted from the load serving as the master server to another load serving as the slave server.
[0008]
[Patent Document 1]
Japanese Patent Laid-Open No. 11-202986
[Problems to be solved by the invention]
In such a power management device, when the router or hub is supplied with power from the commercial power supply, if an abnormality occurs in the commercial power supply, the router is turned off and shuts down to other servers via the network. can not transmit a signal to migrate to, all Masutasa bus than outside and dirty shut down, there is a possibility of damage to data or a hard disk. In such a case, the worst situation was that the server was dirty shut down, leading to a system failure of the slave server.
[0010]
The purpose of the invention is to reliably prevent the connected servers dirty shutdown uninterruptible power supply one, provides a power management how capable power supply, such as during a power failure.
[0011]
[Means for Solving the Problems]
In order to achieve the above object, there is provided a power management method for an uninterruptible power supply system configured by connecting a plurality of servers that receive power supply from the uninterruptible power supply on the same network as the uninterruptible power supply. This power management method always sends an operation confirmation signal from the uninterruptible power supply to each server to recognize the server that is actually operating, and shuts down the OS from the uninterruptible power supply to each server via the network. When sending a command, the server that is actually operating is targeted for sending the OS shutdown command, and the OS shutdown command is repeatedly sent until a response signal to the OS shutdown command is sent back from the server, configured to stop the output of the previous SL uninterruptible power supply after the response signal from the server to the operation confirmation signal has disappeared.
[0012]
According to this power management method, in the OS shutdown of the server using the network, since the completion of the OS shutdown in each server is confirmed after the server is instructed to shut down from the UPS, the dirty OS shutdown is surely avoided. can do.
[0013]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings.
(First embodiment)
1 is a block diagram showing a configuration of an uninterruptible power supply system according to Embodiment 1 of the present invention.
[0014]
In this uninterruptible power supply system, power is supplied from one UPS 1 to two servers 11 and 12, and UPS 1 is configured to back up these two servers 11 and 12 when a power failure occurs in a commercial power source. Has been. The UPS 1 and the servers 11 and 12 can be connected to each other via Ethernet (registered trademark), and are connected to the same LAN here. The hub 2 that mediates this network is also backed up by the UPS 1 in the event of a power failure.
[0015]
Conventionally, the output of the UPS 1 is stopped after a set delay time has elapsed without determining whether the servers 11 and 12 that supply power from the UPS 1 have finished the OS shutdown. . The UPS output stop delay time is set when the server is purchased (or when the system is activated). For this reason, when the system becomes large during use and the OS shutdown time of the server is extended, so-called dirty shutdown occurs, which may destroy data.
[0016]
FIG. 2 is a timing chart showing a shutdown processing procedure according to the first embodiment.
From the UPS 1, a shutdown command C1 is sent to the server 11 set as the OS shutdown target, and a shutdown command C2 is sent to the server 12 (FIGS. 2A and 2C). Thereafter, the UPS 1 sends an operation confirmation signal to each of the servers 11 and 12 in order to know whether the servers 11 and 12 are operating (FIGS. 2E and 2G). As a result, when there is no response to the shutdown command, it can be determined whether the servers 11 and 12 have been stopped in the first place, or whether communication has failed due to traffic on the network.
[0017]
The timing for sending the shutdown commands C1 and C2 from the UPS 1 to the servers 11 and 12 is when a power failure occurs, when a scheduled operation is stopped, or when a UPS stop command is received via the network.
[0018]
Next, the shutdown process procedure when a power failure occurs will be specifically described. When a power failure occurs, the UPS 1 switches from commercial power supply to battery power supply and backs up the servers 11 and 12. When the set time elapses after the power failure, the UPS 1 sends shutdown commands C1 and C2 to the servers 11 and 12 (FIGS. 2A and 2C). Software for performing OS shutdown is installed in each of the servers 11 and 12, and OS shutdown starts immediately upon receiving a shutdown command (FIGS. 2B and 2D). Thereafter, in the conventional system, when the time set in UPS 1 elapses, the output of UPS 1 that is the power source of servers 11 and 12 is stopped regardless of whether the OS shutdown in servers 11 and 12 is completed. However, here, as shown in FIGS. 2 (f) and 2 (h), the servers 11 and 12 send response signals R1 and R2 back to the UPS 1 as responses to the operation confirmation signal, respectively. The power supply to the servers 11 and 12 is stopped when the response signals R1 and R2 with respect to the confirmation signal disappear.
[0019]
That is, if there is a response signal R2 to the operation confirmation signal even when the UPS output scheduled stop time comes, the server 12 does not stop the output of UPS1 (the power supply of the server 2) before the time t0, and the response signal R2 The output of UPS1 is stopped at time t1 when it is confirmed that there is no response. Therefore, the dirty shutdown in the server 12 can be avoided reliably.
[0020]
As described above, in the server OS shutdown using the network, after the UPS instructs the server to shut down, an operation confirmation signal is sent to confirm that the server shutdown has been completed. When the response signal disappears, the completion of OS shutdown in each server is confirmed. Therefore, according to the invention described in the first embodiment, it is possible to provide a power management method for an uninterruptible power supply system that can reliably avoid a dirty OS shutdown of a server.
[0021]
(Second Embodiment)
FIG. 3 is a timing chart showing a shutdown processing procedure according to the second embodiment.
In the first embodiment, it is difficult to retry when the shutdown commands C1 and C2 are not transmitted to the servers 11 and 12. For this reason, when the output is stopped after the set output stop delay time in the UPS 1, there is a possibility that the server in which the shutdown command is not transmitted becomes a dirty shutdown and destroys data.
[0022]
Therefore, the UPS 1 and each of the servers 11 and 12 are configured to always communicate with each other. As shown in FIGS. 3G and 3I , the UPS 1 transfers to the servers 11 and 12 regardless of the shutdown commands C1 and C2. An operation confirmation signal is sent. That is, the response signal R1, R2 to these operation confirmation signal of UPS1 the OS shutdown target server 11, and to recognize the server 12 operating. Therefore, the shutdown command is not sent to the server 11 based on the recognition that the server 11 is stopped.
[0023]
Further, since the UPS 1 recognizes that the server 12 is operating by the response signal R2, as shown in FIG. 3E, a response signal to the shutdown command C2 to the server 12 is sent back to the UPS 1. Yes. Therefore, if there is no response signal, since it performs a retry operation to be sent to repeat the shutdown command C2 to the server 12 until a response signal is sent back, the shutdown to the server 12 can be reliably performed.
[0024]
As described above, according to the second embodiment, by always sending an operation confirmation signal to the servers 11 and 12 to which the shutdown command is sent, the shutdown command is sent only to the server 12 that is actually operating. Send it out. In addition, in the OS shutdown of the servers 11 and 12 from the UPS 1 using the network, it is confirmed that the OS 12 has been shut down in the server 12 because the response of the response signal R2 is lost after the server 12 is instructed to shut down from the UPS 1. Thus, the dirty OS shutdown can be surely avoided by the retry operation.
[0025]
【The invention's effect】
As described above, according to the power management method of the present invention , an operation confirmation signal is sent from the UPS to the server to which the shutdown command is sent, and the shutdown command is sent only to the operating server. Good. In addition, even when the shutdown command is not transmitted to the server, since the dirty OS shutdown is reliably avoided by the retry operation, the reliability of the server OS shutdown using the network can be easily improved.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an uninterruptible power supply system according to Embodiment 1 of the present invention.
FIG. 2 is a timing chart showing a shutdown processing procedure according to the first embodiment.
FIG. 3 is a timing chart showing a shutdown processing procedure according to the second embodiment.
FIG. 4 is a system configuration diagram showing an example of a conventional uninterruptible power supply system.
FIG. 5 is a timing chart for explaining a conventional shutdown processing procedure;
[Explanation of symbols]
1 UPS
2 Hub 11, 12 server

Claims (1)

In a power management method for an uninterruptible power supply system configured by connecting a plurality of server computers that receive power supply from an uninterruptible power supply (hereinafter simply referred to as a server) on the same network as the uninterruptible power supply,
Recognizing servers that are actually operating by sending an operation confirmation signal from the uninterruptible power supply to each server at all times,
When an OS shutdown command is sent from the uninterruptible power supply to each of the servers via the network , a response signal to the OS shutdown command with the server that is actually operating as the sending target of the OS shutdown command Repeatedly sends the OS shutdown command until is sent back from the server,
Power management method characterized by stopping the output of the previous SL uninterruptible power supply after the response signal from the server to the operation confirmation signal has disappeared.

Images