Step 4: Define a Server Machine

Using your local operating procedures, update the z/VM® system directory entries for the new server. Figure 1 shows an example of the z/VM system directory entries. The text following the figure explains the example statements. The user ID of this example server is FPSERV1.

1--> USER FPSERV1 FPSERVPW 32M 32M BG
2--> ACCOUNT acctnum distcode
3--> OPTION MAXCONN 2000 NOMDCFS APPLMON ACCT QUICKDSP SVMSTAT
4--> SHARE REL 1500
5--> MACHINE XC
6--> XCONFIG ADDRSPACE MAXNUMBER 100 TOTSIZE 8192G SHARE
7--> XCONFIG ACCESSLIST ALSIZE 1022
8--> IUCV ALLOW
9--> IUCV *IDENT RESANY GLOBAL
10-> IPL CMS
11-> POSIXOPT SETIDS ALLOW
12-> CONSOLE 009 3215 T userid
13-> SPOOL 00C 2540 READER *
14-> SPOOL 00D 2540 PUNCH A
15-> SPOOL 00E 1403
16-> LINK MAINT 190 190 RR
17-> LINK MAINT 193 193 RR
18-> LINK MAINT 19D 19D RR
19-> MDISK 191 3380 cyl   2 volser1 W
     MDISK 250 3380 cyl  30 volser2 R readpw writepw
     MINIOPT NOMDC
     MDISK 405 3380 cyl  10 volser3 R readpw writepw
     MINIOPT NOMDC
     MDISK 406 3380 cyl  10 volser4 R readpw writepw
     MINIOPT NOMDC
     MDISK 260 3380 cyl  10 volser2 R readpw writepw
     MDISK 310 3380 cyl 200 volser5 R readpw writepw
     MDISK 311 3380 cyl 200 volser3 R readpw writepw
     MDISK 312 3380 cyl 200 volser8 R readpw writepw
     MDISK 313 3380 cyl 200 volser9 R readpw writepw
     MDISK 505 3380 cyl   2 volser6 R readpw writepw
     MINIOPT NOMDC
     MDISK 506 3380 cyl   2 volser7 R readpw writepw
     MINIOPT NOMDC

Use the CMS saved segment name (for example CMS) and device types applicable to your z/VM environment. See z/VM: CP Planning and Administration for more about z/VM system directory control statements.

Other virtual machines should not have LINK control statements in their z/VM system directory entries that refer to file pool minidisks. If such links exists, server processing may not be able to successfully initialize.

Statement 1: USER FPSERV1 FPSERVPW 32M 32M BG

Defines a virtual machine named FPSERV1 with a password FPSERVPW. Neither the machine name nor the password have any significance to the file pool server. Use machine names and passwords of your own choosing.

Although the minimum virtual storage needed for a server is much smaller, 32MB is recommend because that value is adequate for most file pool server machines, and generally prevents you from having to reset the virtual storage size as the use of the file pool grows.

The privilege classes in the example are B and G. It is possible, though not recommended, to define a server machine with only class G. Privilege class B is needed to attach tapes. Class B is needed if you plan to direct control data back ups or the output of security audits to tape. Both backups and security audits are optional operations, but almost everyone backs up the control data, and most of those people direct the backup output to tape. Because file pool is being set up to back up the control data to tape, specify class B.

Statement 2: ACCOUNT acctnum distcode

This statement specifies an account number (acctnum) to which a virtual machine may charge its costs and a distribution code (distcode) to designate where printed output is to go.

Statement 3: OPTION MAXCONN 2000 NOMDCFS APPLMON ACCT QUICKDSP SVMSTAT

The MAXCONN value of the OPTION control statement determines the maximum number of IUCV connections allowed for a virtual machine. The server uses IUCV to communicate with user machines, and uses the DASD Block I/O System Service to access the minidisks in the file pool. The DASD Block I/O System Service also uses IUCV. The maximum MAXCONN value is 65,535—the default is 64. You should specify 2000 to avoid having to frequently re-specify MAXCONN as the use of your file pool grows. A value of 2000 should be adequate for all but very large file pools (that is, file pools with more than 700 logged-on users).

For repository file pool servers, you can estimate the number of IUCV connections that a server needs as follows:

MAXCONN = (SFS_USERS*3) + (BFS_USERS*12) + DISKS

For USERS, substitute the number of users expected during peak system activity. A file pool user is someone who accesses files or directories managed by the repository file pool server. (The sum of SFS_USERS and BFS_USERS should be the same as the nnnnn value used in the USERS startup parameter and File Pool Server Startup Parameters.) For DISKS, substitute the number of minidisks defined for the file pool. Include in this count the number of user data minidisks, catalog minidisks, both file pool repository log minidisks, the control minidisk, and if present, both CRR log minidisks. (This in not applicable to FIFO file pool servers.) Do not include the work minidisk in the count unless the “work disk” is really an SFS directory in another file pool.

For CRR recovery servers, you can estimate the number of IUCV connections a server needs as follows:

MAXCONN = (USERS + RESOURCES + 20)

For USERS, substitute the number of logged-on CRR users, expected during peak system activity. A CRR user is someone who uses:

• Resources that participate in CRR (for example, SFS file pool)
• Protected conversations

If USERS is not known, then substitute the maximum total logged-on CMS users. For RESOURCES, substitute the number of resources participating in CRR (such as SFS file pools). The constant 20, accounts for AVS links and links to other CRR recovery servers and occasional file pool ID links such as for QUERY FILEPOOL REPORT.

For dedicated FIFO servers:

MAXCONN = (USERS*2) + SERVERS

For USERS substitute the number of users of the FIFO server expected during peak activity. For SERVERS substitute the approximate number of servers that designate this FIFO file pool in their file pool startup parameter (FIFO (filepoolid)).

Server processing sets aside virtual storage for each potential connection. Thus, setting MAXCONN too high may cause virtual storage to be exhausted in the server machine.

Repository file pool servers use NOMDCFS to allow the server machine to use minidisk caching at a rate not limited by the Fair Share Limit. CRR recovery servers should not use NOMDCFS, because the CRR minidisks do not benefit from minidisk caching.

The APPLMON operand lets the server machine generate information needed for performance monitoring. ¹

The ACCT operand is required if you intend to use the file pool accounting facilities for this server. If you do not intend to use the accounting facilities, you can omit ACCT.

The QUICKDSP operand causes the server to be added to the dispatch list immediately when it has work to do, without waiting in the eligible list.

The SVMSTAT operand specifies this server is a service virtual machine, which causes the server's monitor statistics to be reported separately from end-user virtual machines.

Statement 4: SHARE REL 1500

This statement places the server machines in a more favorable position in the z/VM dispatch queue. Use this statement for all server machines.

Statement 5: MACHINE XC

This statement specifies the type of architecture a virtual machine simulates. XC designates ESA/XC or z/XC architecture; ESA designates ESA/390 architecture.

For SFS file pool servers, designate XC if you want to exploit data spaces. CRR recovery servers, BFS, and FIFO do not exploit data spaces, and, therefore, should be set to ESA.

Statement 6: XCONFIG ADDRSPACE MAXNUMBER 100 TOTSIZE 8192G SHARE

This statement authorizes the virtual machine to create data spaces and sets limits on the number and size of those data spaces. As a guideline, specify the maximum number of data spaces you want as five times the number of directories you plan to make eligible for data spaces. You need to specify this statement if you intend to use data spaces. See Configuring Your Server for Data Spaces for more information about the XCONFIG statements.

Statement 7: XCONFIG ACCESSLIST ALSIZE 1022

This statement sets up space for the server's host access list. You need to specify this statement if you intend to use data spaces. See Configuring Your Server for Data Spaces for more information about the XCONFIG statements.

Statement 8: IUCV ALLOW

This statement lets any other virtual machine establish a communication path with the server machine.

For most server machines, you should specify IUCV ALLOW as shown. If you prefer, however, you can use different IUCV control statements to control the communications between user machines and the server. Server processing uses APPC/VM to communicate with users. See z/VM: Connectivity for more about inter-machine communications.

Statement 9: IUCV *IDENT RESANY GLOBAL

This IUCV *IDENT statement authorizes the virtual machine to identify itself as a resource owner.

RESANY must be specified because server processing lets you change the resource name when you start it (using the FILESERV startup parameter named FILEPOOLID).

GLOBAL should be specified in most cases.

For dedicated CRR recovery servers, regardless of whether the file pool ID has a VMSYS prefix, you must always specify GLOBAL.

For repository and FIFO servers, if the file pool ID does not have a VMSYS prefix, you must always specify GLOBAL. If the file pool ID does have a VMSYS prefix, you can specify LOCAL or GLOBAL (LOCAL is sufficient, because in this case, the server always identifies the file pool ID TPN as local).

For more information on how file pool ID TPNs are identified, see the startup parameter FILEPOOLID.

Statement 10: IPL CMS

This statement causes CMS to be loaded into the server machine when it is logged on. If you prefer, you can use the appropriate saved segment name for your installation. IPL ZCMS can also be an appropriate choice.

Do not specify the AUTOCR initialization parameter for CMS. To enter some of the dedicated maintenance mode commands for the server, it may be necessary to avoid executing the server's PROFILE EXEC. This is not possible if you specify AUTOCR.

Statement 11: POSIXOPT SETIDS ALLOW

This statement allows the BFS server to have CP change its POSIX security value in support of opening executable files.

Statement 12: CONSOLE 009 3215 T userid

During usual operation, you will be running the server machine in disconnected mode. So, you should specify a secondary user ID. This is known as Single Console Image Facility (SCIF). If you will be acting as the administrator or operator of the file pool, you should specify your own user ID. If you are generating the file pool for a secondary user ID who is to receive console output, specify that person's user ID.

Statement 13: SPOOL 00C 2540 READER *

This statement defines at 00C a spooled 2540 reader that can read any class of spool file.

Statement 14: SPOOL 00D 2540 PUNCH A

This statement defines at 00D a spooled 2540 punch that creates only class A spool files.

Statement 15: SPOOL 00E 1403

This statement defines at 00E a spooled 1403 printer that processes class A spool files (when the spooling class is omitted, it defaults to A).

Statement 16: LINK MAINT 190 190 RR

This is a link statement for access to CMS.

Statement 17: LINK MAINT 193 193 RR

This statement is necessary to make the MAINT 193 minidisk available to the server machine. Some of the files a server needs (such as the FILESERV EXEC) are located on this minidisk. This LINK statement is, therefore, required.

Statement 18: LINK MAINT 19D 19D RR

This is a link statement for the Help Facility.

Statement 19: MDISK 191 3380 cyl 2 volser1 W

This is the MDISK statement for the read/write work disk on which the POOLDEF file will reside. Use virtual device number 191 and access mode W as shown. In this example, a device type of 3380 that consists of two cylinders is used. You also need to supply the starting cylinder number and the volume serial number.

Remaining MDISK and MINIOPT NOMDC statements:

MDISK 250 3380 cyl  30 volser2 R readpw writepw
MINIOPT NOMDC
MDISK 405 3380 cyl  10 volser3 R readpw writepw
MINIOPT NOMDC
MDISK 406 3380 cyl  10 volser4 R readpw writepw
MINIOPT NOMDC
MDISK 260 3380 cyl  10 volser2 R readpw writepw
MDISK 310 3380 cyl 200 volser5 R readpw writepw
MDISK 311 3380 cyl 200 volser3 R readpw writepw
MDISK 312 3380 cyl 200 volser8 R readpw writepw
MDISK 313 3380 cyl 200 volser9 R readpw writepw
MDISK 505 3380 cyl   2 volser6 R readpw writepw
MINIOPT NOMDC
MDISK 506 3380 cyl   2 volser7 R readpw writepw
MINIOPT NOMDC

In the preceding MDISK statements examples, the supplied virtual device numbers (191, 250, 405, 406, and others) and also the device type (3380) are shown. The MDISK statements shown are for count-key-data devices, so you must supply starting cylinder number (cyl) followed by the number of cylinders (the number of cylinders was supplied in the example). For FB-512 devices, specify the appropriate block relocation factor (blkr) and number of blocks (blks). You also must supply the volume serial number (volser1 through volser8) of the real DASD volume on which the minidisk resides. Use access mode R and then supply a read password (readpw) and a write password (writepw) for each minidisk.

Note:

1. Set the volume serial numbers (volser) as appropriate for your DASD volumes.
2. The MDISK control statements for the file pool must include both a read and a write password. Also, it is strongly recommended the minidisks have an access mode of R. Specify your own read and write passwords for readpw and writepw.
3. Except for the 191 minidisk, any virtual device numbers can be used in any order. Server processing does not rely on any particular device numbers. In this example, virtual device numbers 250, 405, 406, 260, 310, 311, 312, 313, 505, and 506 were chosen arbitrarily.
4. When you enter the FILESERV GENERATE command (see Step 8: Generate the File Pool), you must be prepared to supply the virtual device numbers of the minidisks you used in the MDISK statement. You should write down your choices of virtual device numbers along with how they are to be used (for example, control minidisk, file pool repository log minidisk 1, and so on).
5. FBA minidisks must start and end on 4KB boundaries or data spaces will not be created by this file pool server, even for data that is on aligned FBA or CKD DASD.
6. In this example, the minidisk having a virtual device number of 191 is the work minidisk. The 250 minidisk is the file pool control minidisk. It is defined for an IBM 3380.

   Minidisks 405 and 406 are the file pool repository logs. Notice the file pool repository logs are defined on different volumes (volser3 and volser4). They are both the same size and are defined on the same device type, which again is an IBM 3380.

   Minidisk 260 is the initial catalog minidisk. It is on the same volume as the control minidisk (250) because both are recovered as a unit.

   Minidisks 310, 311, 312, and 313 are for user data.

   Minidisks 505 and 506 are for the CRR logs. Notice the CRR logs are defined on different volumes (volser6 and volser7). They are both the same size and are defined on the same device type, which again is an IBM 3380.

7. In the preceding example of MDISK control statements, a server is shown performing both repository file pool server functions and CRR recovery server functions.

   If this was a repository file pool server, you would not have minidisks for the CRR logs (505 and 506).

   If this was a dedicated CRR recovery or FIFO server, the sizes of the control minidisk (250), file pool repository log minidisks (405 and 406), catalog minidisk (260), and the user minidisk (310) would be much smaller. For example, each minidisk would probably be only 1 to 2 cylinders. Also, you would not need the other user minidisks (311, 312, and 313), because CRR recovery or FIFO server activity is not intended for file pool repository use. You would still have the CRR log minidisks (505 and 506).

Use the MINIOPT NOMDC NOCACHE control statement to inhibit expanded storage caching; omit the NOCACHE option (MINIOPT NOMDC) when you are using DASD fast-write. One reason to do this is to improve system performance by not caching the file pool repository log and CRR log minidisks. Table 1 lists the uses of the MINIOPT control statement.

---

Table 1. Uses of MINIOPT NOMDC and MINIOPT NOMDC NOCACHE

Minidisk	Repository File Pool	Dedicated CRR or FIFO File Pool
Control	Yes	Yes
Log 1 and Log 2	Yes	Yes
Catalog	No	Yes
User	No	Yes
CRR Log 1 and Log 2	N/A	Yes

---