JP2011187021A - Blade server - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a compact blade server that analyzes a failure in a storage medium. <P>SOLUTION: A storage area of a storage device mounted on each of a plurality of blades comprises an individual area, a blade common area and an auxiliary area. Information data is written to or read from a RAID area configured by virtually connecting the blade common area of the storage device in each blade. A log file and/or a core file as a failure analysis file are/is written into the blade common area of the storage device in each blade. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a server that provides various types of information data, and more particularly to a blade server that includes a plurality of blades each equipped with a storage device for storing information data.

  Currently, blade servers that meet demands such as space saving, operational management efficiency, and maintenance cost reduction are attracting attention as servers that provide various information data on an information network. The blade server is constructed from a plurality of modules called blades each equipped with a processor, a memory, a network connection interface, a storage device for storing various information data, and the like.

  As such a blade server, a blade server in which a plurality of disk drive devices as storage devices are mounted on each blade is known (see, for example, the blade 200 and the disk drive 218 in FIG. 2 of Patent Document 1).

  By the way, when considering miniaturization of the entire device, it is preferable that the number of disk drive devices mounted on each blade is one.

  However, if only one disk drive device is mounted on one blade, when a failure occurs in the disk drive device, it becomes impossible to read a log file held by the blade itself. Therefore, there arises a problem that the analysis cannot be performed after the failure of the disk drive device occurs, or the past operation state cannot be referred to.

Special table 2007-526527

  An object of the present invention is to provide a blade server that can analyze a storage medium when a failure occurs while realizing a reduction in size.

  The blade server according to the present invention reads a plurality of blades each having a storage device mounted thereon, a blade rack in which each of the blades is mounted, and the storage device mounted on each of the blades. Or a control unit that performs write control, wherein each storage area of the storage device includes an individual area, a blade common area, and an increase / decrease area, and the control unit One area obtained by virtually linking the blade common areas of each of the storage devices mounted on each of the storage apparatuses is regarded as a raid area, and information data write or read access is performed on the raid area. An analysis file is written to the blade common area in each storage device.

  In the present invention, information data write or read access is executed to a raid area formed by virtually linking a part of each storage device (blade common area) mounted on each blade. Further, when a failure occurs in the storage device, a log file and / or a core file are written in the blade common area of each storage device for each blade as information necessary for performing the failure analysis. As a result, when a failure occurs in a storage device installed in one of the blades, log files or core files stored in the blade common area of the storage devices installed in other blades By referring to the failure, the failure analysis can be performed. Therefore, according to the present invention, even when only one storage device is mounted on each blade in order to reduce the size of the entire device, an analysis is performed when a failure occurs in this one storage device. It becomes possible.

It is a figure which shows an example of a structure of the blade server by this invention. It is a figure which shows the read-write access control flow implemented with CPU2 mounted in each blade BL. It is a figure which shows an example of the state of each storage area of HDD4 ₁ -HDD4 ₃ mounted in each of the blades BL1 to BL3 shown in FIG. It is a figure which shows the flow of the raid area | region access control implemented in the raid controller. It is a figure which shows the state of the storage area of each HDD4 just before adding blade BL4. It is a figure which shows the flow of the blade expansion control implemented in the RAID controller 100 according to the expansion of a blade. It is a figure which shows the state of the storage area of each HDD4 in the expansion 1st step of blade BL4. It is a figure which shows the state of the storage area of each HDD4 in the expansion 2nd step of blade BL4. It is a figure which shows the state of the storage area of each HDD4 in the expansion last stage of blade BL4. It is a figure which shows the flow of the blade reduction control implemented in the raid controller 100 according to the reduction of a blade. It is a figure which shows the state of the storage area of each HDD4 in the reduction | decrease 1st step of blade BL1. It is a figure which shows the state of the storage area of each HDD4 in the reduction | decrease 2nd step of blade BL1. It is a figure which shows the state of the storage area of each HDD4 in the reduction | decrease 3rd step of blade BL1. It is a figure which shows the state of the storage area of each HDD4 in the reduction | decrease final stage of blade BL1.

  The storage area of the storage device mounted on each of the plurality of blades includes an individual area, a blade common area, and an increase / decrease area, and is formed by virtually linking the blade common areas of the storage devices for each blade. Information data writing or reading access to the raid area is executed, and a log file and / or core file as a failure analysis file is written to the blade common area of each storage device for each blade.

  FIG. 1 is a diagram showing an example of the configuration of a blade server according to the present invention.

  The blade server shown in FIG. 1 includes an inter-blade communication path TL, a blade rack BK, blades BL1 to BL3, and a RAID (Redundant Array of Inexpensive Disks) controller 100 (hereinafter referred to as a raid controller 100). In the example shown in FIG. 1, blades BL1 to BL3 are mounted in SL1 to SL3 among the four slots SL1 to SL4 of the blade rack BK, respectively, and the slot SL4 is a slot for blade expansion. . The blade rack BK supplies a blade addition detection signal to the RAID controller 100 indicating that the blade has been added when the blade is installed in the slot SL4.

  A RAID controller 100 and three blades BL1 to BL3 are connected to the inter-blade communication path TL. Each of the blades BL1 to BL3 is provided with a communication interface unit 1, a CPU (Central Processing Unit) 2, a disk controller 3, and a single hard disk drive device 4 (hereinafter referred to as HDD 4).

  The communication interface 1 sends the information data read from the HDD 4 as a storage device to the inter-blade communication path TL, and takes in the information data on the inter-blade communication path TL and supplies it to the HDD 4. Further, the communication interface 1 takes in the RAID reading access signal AS sent from the RAID controller 100 from the inter-blade communication path TL, and supplies it to the CPU 2.

The CPU 2 supplies a read command signal for reading information data stored in the HDD 4 to the disk controller 3 in accordance with a program (described later) stored in a built-in memory (not shown). Further, the CPU 2 supplies a write command signal for causing the HDD 4 to write information data supplied from the inter-blade communication path TL to the disk controller 3 in accordance with such a program. Further, when a later-described read / write end signal EN is supplied from the disk controller 3, the CPU 2 transmits this to the RAID controller 100 via the communication interface 1 and the inter-blade communication path TL. Further, the CPU 2 transmits a raid reading request signal R _REQ for requesting the raid reading access signal AS to the raid controller 100 via the communication interface 1 and the inter-blade communication path TL according to the program.

  When a read command signal is supplied from the CPU 2, the disk controller 3 supplies a read signal for reading stored information data to the HDD 4, while when a write command signal is supplied, the communication interface 1, a write signal for writing information data supplied from the inter-blade communication path TL is supplied to the HDD 4. Further, when the writing or reading process in the HDD 4 is completed, the disk controller 3 supplies a read / write end signal EN indicating that to the CPU 2.

  When a read signal is supplied from the disk controller 3, the HDD 4 reads information data and supplies the read information data to the communication interface 1. Further, when a write signal is supplied from the disk controller 3, the HDD 4 writes the information data supplied from the communication interface 1 into the storage area. The storage area of the HDD 4 is divided in advance into three areas: an individual area INA, a blade common area COA, and an increase / decrease reserve area AUA. The individual area INA stores a program OS related to an operating system that controls the blade BL in which the HDD 4 is mounted.

The RAID controller 100 writes or writes information data to each of the blades BL1 to BL3 in order to operate the HDD 4 mounted on each of the blades BL1 to BL3 as one virtual file storage device. Raid reading access signals AS1 to AS3 for executing reading are generated. Hereinafter, the HDD 4 mounted on the blade BL1 is referred to as HDD 4 ₁ , the HDD 4 mounted on the blade BL 2 is referred to as HDD 4 ₂ , and the HDD 4 mounted on the blade BL ₃ is referred to as HDD 4 ₃ . Here, the RAID controller 100 adopts RAID 5 of RAID 0 to 5 known as RAID levels, for example, divides one information data by the number of blades, and each is mounted on each blade. A RAID reading access signal AS to be read / written to be distributed to the HDDs 4 is generated. At this time, the RAID controller 100 creates parity data for information data restoration based on the information data to be written to the HDD 4 of each blade, and distributes a plurality of parity blocks composed of the parity data in the information data series. And insert. That is, the RAID controller 100 reads or writes the first divided data (including the parity block) of the first to third divided data obtained by dividing the information data series into three. AS1, a RAID reading access signal AS2 from which the second divided data (including the parity block) should be read or written, and a RAID reading access signal AS3 from which the third divided data (including the parity block) should be read or written Generate. Then, the raid controller 100 transmits a raid reading access signal AS1 to the blade BL1, a raid reading access signal AS2 to the blade BL2, and a raid reading access signal AS3 to the blade BL3. As a result, the information data including the parity block as described above is written in each of the HDDs 4 ₁ to 4 ₃ . With such a writing mode, even if _{one of the} HDDs 4 ₁ to 4 ₃ mounted on each of the blades BL 1 to BL ₃ fails, the information data and parity blocks stored in the remaining HDD 4 On the basis of the above, it is possible to restore the information data stored in the failed HDD 4. The raid controller 100 incorporates a management memory in which management information for managing each storage area of each of the blades BL1 to BL3 is stored. As management information, in addition to information indicating individual area INA, blade common area COA, increase / decrease spare area AUA, information indicating storage capacity, etc., the storage areas of each of the plurality of HDDs 4 are virtually linked. There is raid management information that represents an address for each blade when the one area (raid area) is set. At this time, the management memory in the initial state stores raid management information in which the blade common areas COA of the HDDs 4 ₁ to 4 ₃ are virtually connected to each other as shown in FIG.

  Next, an outline of the data read / write operation in the blade server shown in FIG. 1 will be described.

  FIG. 2 shows a case where a write or read request (hereinafter referred to as an access request) is made during execution of a main program (not shown) in one of the CPUs 2 mounted on the blades BL1 to BL3. It is a figure which shows the flow of the read-write access control performed.

  In FIG. 2, first, the CPU 2 determines whether the access request is issued from the RAID controller 100, that is, whether a log access request is issued from the RAID controller 100, or whether the main request executed by the CPU 2 is executed. It is determined whether the access request is obtained by the program (step S1). The log access request from the RAID controller 100 refers to a log file indicating a history of various accesses made to the blade BL to which the CPU 2 belongs and / or information data that was used immediately before the HDD 4 failed. This is for requesting writing or reading to the indicated core file. If it is determined in step S1 that the access request is from the RAID controller 100, the CPU 2 writes the log file and / or core file to the blade common area COA or the additional spare area AUA of the HDD 4 or A write command signal or a read command signal to be read is supplied to the disk controller 3 (step S2). By executing step S2, the HDD 4 writes or reads the log file and / or core file indicated by the access request to the blade common area COA or the expansion spare area AUA. At this time, when the HDD 4 finishes accessing the blade common area COA or the expansion spare area AUA, the disk controller 3 supplies the CPU 2 with a read / write end signal EN. After execution of step S2, the CPU 2 repeatedly determines whether or not the read / write end signal EN has been supplied (step S3). When it is determined that the read / write end signal EN has been supplied, this read / write access control routine Return to the execution of the main program.

  If it is determined in step S1 that the access is not from the RAID controller 100, the CPU 2 indicates that the above access request indicates access to the individual area INA of the HDD 4 mounted on the blade BL to which the CPU 2 belongs. Is determined (step S4). If it is determined in step S4 that the request is an access request to the individual area INA, the CPU 2 reads or writes information data from or to the individual area INA of the HDD 4 according to the access request to the disk controller 3. A write command signal or a read command signal to be executed is supplied (step S5). By executing step S5, the HDD 4 writes or reads information data to or from the individual area INA. At this time, when the access to the individual area INA by the HDD 4 is completed, the disk controller 3 supplies a read / write end signal EN to the CPU 2. After executing step S5, the CPU 2 repeatedly determines whether or not the read / write end signal EN has been supplied (step S6). If it is determined that the read / write end signal EN has been supplied, this read / write access control routine is executed. Return to the execution of the main program.

On the other hand, if it is determined in step S4 that the access request does not indicate an access to the individual area INA of the HDD 4, the CPU 2 reads the raid reading request signal R _REQ for requesting the raid reading access signal AS. Is transmitted to the RAID controller 100 via the communication interface 1 and the inter-blade communication path TL (step S7). Here, when the raid controller 100 receives the raid reading request signal _RREQ , the raid controller 100 executes a raid area access control routine as shown in FIG.

In FIG. 4, first, the raid controller 100 generates the raid reading / accessing signals AS1 to AS3 as described above based on the raid management information stored in the management memory (step S21). Next, the raid controller 100 transmits the raid reading access signal AS1 to the blade BL1, the raid reading access signal AS2 to the blade BL2, and the raid reading access signal AS3 to the blade BL3 via the inter-blade communication path TL (step S22). ). By executing step S22, each of the blades BL1 to BL3 performs information data read or write access to the area designated as the raid area RA in the storage area of the HDD 4 mounted therein. At this time, the blade BL that has finished reading or writing information data to the HDD 4 transmits a read / write end signal EN to the RAID controller 100 via the inter-blade communication path TL. During this time, the RAID controller 100 determines whether all the read / write end signals EN transmitted from the blades BL1 to BL3 have been received, and determines that all the read / write end signals EN transmitted from the blades BL1 to BL3 have been received. It repeats until it does (step S23). When it is determined in step S23 that the read / write end signal EN has been supplied from all the blades BL1 to BL3, the raid controller 100 indicates that the access to the raid area RA as shown in FIG. 3 has been completed. the R _END, through the inter-blade communication path TL, and transmits to the blade BL, which is the transmission source of the as mentioned above laid reading request signal _{R REQ.}

During this time, CPU 2 mounted on which is the source blade BL is repeated until it is determined the decision of whether or not it has received the raid access completion signal R _END has been received the raid access completion signal R _END ( Step S8 in FIG. If it is determined in step S8 that the RAID access completion signal R _END has been received, the CPU 2 exits the read / write access control routine shown in FIG. 2 and returns to the execution of the main program.

As described above, in the blade server shown in FIG. 1, a part of the storage area (blade common area INA) in the storage area of the HDD 4 mounted on each of the blades BL1 to BL3 is virtually as shown in FIG. In this way, one raid area RA is constructed, and information data is written or read with respect to the raid area RA area. That is, by using the raid area RA, each of the divided data obtained by dividing the series of information data series by the number of blades BL (three) is assigned to the HDD4 _{1 of} the blade BL1, the HDD4 _{2 of} the blade BL2, and the blade BL3. so that to write by respectively dispersing the HDD 4 _3. At this time, in such a blade server, a log file and / or a core file are used as information necessary for performing a failure analysis when a failure occurs in the HDD 4, and a blade common area which is an area in charge of each blade in the RAID area RA. The data is written in the COA or expansion spare area AUA. As a result, even if _{one of the} HDDs 4 ₁ to 4 ₃ mounted on the blades BL ₁ to BL ₃ fails, the log stored in the blade common area COA or the additional spare area AUA of the other HDD 4 Failure analysis is possible by referring to the file or the core file.

  Next, an operation performed when a blade is added will be described by taking as an example a case where the blade BL4 is added to the blade server shown in FIG.

As shown in FIG. 5, in each individual area INA of each of the HDDs 4 ₁ to 4 ₃ mounted on the installed blades BL ₁ to BL _3, a program OS related to the operating system for each blade BL is stored in advance. . Further, as shown in FIG. 5, the management memory of the RAID controller 100 stores RAID management information in which one area obtained by virtually connecting the blade common areas COA of the HDDs 4 ₁ to 4 ₃ is defined as the RAID area RA1. ing. In this raid area RA, log files LF1 to LF3 and / or core files CF1 to CF3 for each blade BL are stored in advance. At this time, the total storage capacity of each of the increase / decrease spare areas AUA of the HDDs 4 ₁ to 4 ₃ and the storage capacity of the RAID area RA are the same. Further, the individual area INA of ₄ HDD 4 mounted on the blade BL4 for expansion and program OS about the operating system of the blade BL4 is stored in advance, the HDD 4 ₄ blades common area COA or decrease設用spare area AUA Is assumed to store the log file LF4 and / or the core file CF4 of the blade BL4.

  FIG. 6 is a diagram showing a blade addition control flow executed by the raid controller 100 in response to a blade addition detection signal sent from the blade rack BK when the blade BL4 is mounted in the empty slot SL4 of the blade rack BK.

In FIG. 6, first, the RAID controller 100 regards each of the increase / decrease spare areas AUA for the HDDs 4 ₁ to 4 ₄ as blade common areas COA as shown in FIG. The raid management information stored in the management memory is updated so as to make the additional raid area RA2 (step S61).

Next, the raid controller 100 generates raid reading access signals AS1 to AS3 to copy all the contents stored in the raid area RA1 to the additional raid area RA2, and transmits them to the blades BL1 to BL3 (step S62). ). As a result, each of the HDDs 4 ₁ to 4 ₃ mounted on the blades BL ₁ to BL ₃ stores the log files LF 1 to LF 3 and the core files CF ₁ to CF ₃ stored in the raid area RA 1 as shown in FIG. Various information data included is copied to the extension raid area RA2 and synchronized.

Then, the RAID controller 100, with respect to the blade BL4, should the log file LF4 and / or core file CF4 stored in HDD 4 ₄ blade supply area COA of the blade BL4, it is copied to an expansion raid area RA2 log An access request is made (step S63). In response to the log access request, HDD 4 ₄ blades BL4, as shown in FIG. 8, the log file LF4 and / or core file CF4 is stored in the blade supply area COA, copy the expansion raid area RA2 Synchronize.

Then, the RAID controller 100, the RAID area RA1 shown in FIG. 8, the cut into HDD 4 ₁ to 4 _3, respectively for each increase or decrease設用reserve area AUA 9, of such HDD 4 ₄ 8 blades In order to change the common area COA to the increase / decrease reserved area AUA, the management information stored in the management memory is rewritten (step S64).

  The execution of steps S61 to S63 as described above completes the addition of the blade BL4. As a result, as shown in FIG. 9, an additional RAID area RA2 including the blade common area COA of the blade BL4 is formed in the blade server.

  On the other hand, when a blade is removed from the blade server shown in FIG. 1, the following processing is performed.

  FIG. 10 shows that the user performs blade reduction input for designating the blade BL1 as a blade to be removed from the blades BL1 to BL4 shown in FIG. 9 at the operation unit (not shown) of the RAID controller 100. 5 is a diagram showing a blade reduction control flow executed by the raid controller 100 when the

In FIG. 10, first, the raid controller 100, as shown in FIG. 9, shows the spare area AUA for increasing / decreasing each of the HDDs 4 ₂ to ₄ 4 mounted on the blades BL 2 to BL 4 excluding the blade BL 1 to be removed. As shown in FIG. 11, the raid management information stored in the management memory is updated so as to be the reduced raid area RA3 which is changed to the blade common area COA and regarded as one area virtually connected to each other (step S101). .

Next, the raid controller 100 generates raid reading access signals AS1 to AS4 to copy all the contents stored in the additional raid area RA2 shown in FIG. 11 to the reduced raid area RA3, and sends them to the blades BL1 to BL4, respectively. Transmit (step S102). As a result, each of the HDDs 4 ₁ to ₄ 4 mounted on the blades BL ₁ to BL 4 has the log files LF 1 to LF 4 and / or core files CF ₁ to CF 4 stored in the additional RAID area RA 2 as shown in FIG. Are copied to the reduced raid area RA3 and synchronized.

Next, the raid controller 100 manages the management information stored in the management memory so as to divide the additional raid area RA2 shown in FIG. 12 into the spare area AUA for increase / decrease for each of the HDDs 4 ₁ to 4 _{4 as} shown in FIG. Is rewritten (step S103).

Then, the RAID controller 100 is stored in degrowth laid area RA3, the log file LF1 and / or core files CF1 corresponding to the blade BL1, HDD 4 ₁ increase or decrease設用spare such blades BL1 shown in FIG. 12 The raid reading access signals AS2 to AS4 to be moved to the area AUA are transmitted to the blades BL2 to BL4 and a log access request is made to the blade BL1 (step S104). Thus, as shown in FIG. 13, the HDD 4 ₁ increase or decrease設用spare area AUA blade BL1 log file LF1 and / or core file CF1 is written.

  After completion of the above steps S101 to S104, the user removes the blade BL1 from the blade rack BK as shown in FIG. Thereby, the raid area (RA3) is reconstructed by the three blades BL2 to BL4 excluding the blade BL1.

  Therefore, according to the blade server shown in FIG. 1, when a new blade is attached or removed, a part of the HDD (blade common area COA) in the HDD mounted on each blade is automatically added. A raid area formed by virtually connecting the two is constructed.

2 CPU
3 Disk controller 4 HDD
100 raid controller

Claims (6)

A plurality of blades each mounted with a storage device, a blade rack on which each of the blades is mounted, and a control for performing read or write control on the storage device mounted on each of the blades A blade server comprising:
The storage area of each storage device includes an individual area, a blade common area, and an increase / decrease area,
The control unit regards one area virtually connecting the blade common areas of each storage device mounted on each of the blades as a raid area and writes information data to the raid area or A blade server characterized by executing a read access and writing a failure analysis file to the blade common area in each storage device.   The blade server according to claim 1, wherein the failure analysis file is a log file and / or a core file. When the blade rack detects an additional blade, the control unit
Constructing an additional raid area by virtually connecting the increase / decrease area of the storage device of each of the blades including the additional blade; and
Copying the content stored in the raid area to the additional raid area;
Blade addition control including: changing the blade common area of the storage device mounted on the expansion blade to the blade common area, and dividing the raid area into the increase / decrease area for each storage device The blade server according to claim 1, wherein the blade server is executed. The control unit, when a blade reduction input is made,
Constructing a reduced raid area by virtually connecting the increase / decrease area of each of the storage devices mounted on each of the blades excluding the blade to be removed; and
Copying the content stored in the raid area to the reduced raid area;
The history information corresponding to the storage device mounted on the removal target blade is extracted from the reduced raid area, and the history information corresponding to the storage device mounted on the removal target blade is extracted from the history information corresponding to the storage device. The blade server according to claim 1, wherein blade reduction control including a step of writing to the area is executed.   The blade server according to claim 1, wherein each storage device is mounted on each of the blades.   The blade server according to claim 1, wherein a program related to an operating system for each blade is stored in the individual area in advance.

Images