The Latest in z/os UNIX and zfs

Transcription

1 The Latest in z/os UNIX and zfs Marna WALLE, Member of the IBM Academy of Technology z/os System Installation IBM Z Systems, Poughkeepsie NY USA Thanks for material provided by Vivian Morabito, zfs Function Test, and Bill Schoen, z/os UNIX Development.

2 Agenda zfs: Encryption and Compression Shrink SMF Recording Implicit Format on mount Online Salvage (verify / repair) Supplemental: Dynamic Change Aggregate Attributes (chaggr), Miscellaneous updates, and zfs V2R3 Performance Measurements z/os UNIX: HFS to zfs utility Automatic unmount of version file system in shared file system Maintaining a single configuration file across the sysplex Assigning default umask values to all z/os UNIX users Copyright IBM Corporation Materials may not be reproduced in whole or in part without the prior written permission of IBM. 1 1

3 zfs Encryption AND compression V2R3 Encryption & Compression Support for zfs New and Existing filesystems can be encrypted and / or compressed User decides which operation, and in what order. HOWEVER, if you want to do both, compression before encryption is highly recommended. After a filesystem is fully encrypted or compressed, additional new entries will automatically be encrypted or compressed Copyright IBM Corporation Materials may not be reproduced in whole or in part without the prior written permission of IBM. 2 2

4 zfs Encryption AND compression V2R3 Encryption & Compression Support for zfs Encrypting or compressing an existing filesystem is a long-running administrative command Application access is fully allowed to the filesystem during the operation. Progress may be monitored with FSINFO Copyright IBM Corporation Materials may not be reproduced in whole or in part without the prior written permission of IBM. 3 3

5 zfs Encryption AND compression zfs V2R3: New long running admin commands New thread pool created for long running commands Foreground threads handle overall long running operation(s) Background threads used by foreground thread to allow for parallelism new configuration option long_cmd_threads long_cmd_threads=foreground, background (each value can be 1-64, default is 8,24) Long running commands require an available foreground thread 4 4

6 zfs Encryption AND compression zfs V2R3: New long running admin commands Background tasks on the zfs owning system process every object in the filesystem. Try not to use long running commands (encrypt, compress, shrink) during periods of high activity. Best to use them during off shift time periods 5 5

7 zfs Encryption / Compression zfs Encryption & Compression are both V2R3 ONLY If there are V2R1 or V2R2 members in the sysplex: Encryption or Compression may not be initiated. If you mount an encrypted or compressed filesystem on a V2R3 member in a sysplex with V2R1 or V2R2 members, zfs will function ship to the downlevel members (significant performance impact) Wait until you are fully on V2R3 on all systems (with no intent to go back) before using zfs encryption and compression services! 6 6

8 zfs Encryption / Compression commands: Decryption & Decompression are supported Allows you to back out, if pre-v2r3 members need to be brought back into the sysplex 7 7

9 zfs Encryption / Compression Consider pairing encryption with compression. If the compression is done first the amount of data to be encrypted is smaller which may improve performance. zfs does not perform a dual compression and encryption: user directs the actions separately. You cannot encrypt or decrypt an aggregate that is partially compressed or partially decompressed. You cannot compress / decompress an aggregate that is partially encrypted or partially decrypted. 8 8

10 zfs Encryption Utilizes DFSMS Encryption for VSAM datasets uses ICSF* to perform encryption and decryption and to manage cryptographic keys. Userid associated with zfs need access to those keys. ICSF configuration is beyond the scope of this zfs-specific presentation. Ensure that ICSF is running early in the IPL (COMMNDxx, not automation, S CSF,SUB=MSTR) before you exploit DFSMS data set encryption. (zfs relaxes the DFSMS rule that the data set be SMS-managed and extended format.) zfs encrypts: file contents, security information, ACLs (Access Control Lists) symbolic link contents zfs does NOT encrypt the zfs cache * ICSF: z/os Integrated Cryptographic Service Facility (ICSF) 9 9

11 zfs Encryption Only V1.5 filesystems may be encrypted Filesystem must be mounted R/W new configuration option: EDC_BUFFER_POOL Specifies size of real storage permanently fixed and reserved for encryption & compression I/O Default: 32 M for zfs PFS, setting may range from 1 M to 1 G ICSF must be active to encrypt Ensure the PTF for APAR OA54005 is installed

12 zfs Encryption Key Label Encryption required a key label encryption key labels to identify encryption keys to be used to encrypt selected data sets. The specified key label and encryption key must exist in the ICSF key repository (CKDS) 11 11

13 zfs Encryption Key Label Key labels on a filesystem cannot be changed after you assign them Even if filesystem has not yet been encrypted The assignment of a key label does not encrypt a filesystem it makes it eligible for encryption Once a filesystem has a key label, it can never be owned by a member at a level lower than V2R

14 zfs Encryption Creating a NEW encrypted filesystem 1. Define filesystem 2. Assign a key label 3. Format for encryption Data will automatically encrypted on write 13 13

15 zfs Encryption New Encrypted filesystem: DEFINE and assign IDCAMS DEFINE CLUSTER (from JCL) Must use new ZFS keyword KEYLABEL(keylabel) keyword zfsadm define (from the shell) keylabel keylabel 14 14

16 zfs Encryption IDCAMS DEFINE CLUSTER and assign example //ZDEFKEY JOB,'DEFINE WITH KEYLABEL', // MSGCLASS=H, // CLASS=A, // TIME=(1440),MSGLEVEL=(1,1) //DEFINE EXEC PGM=IDCAMS //SYSPRINT DD SYSOUT=H //SYSUDUMP DD SYSOUT=H //AMSDUMP DD SYSOUT=H //DASD0 DD DISP=OLD,UNIT=3390,VOL=SER=POSIX1 //SYSIN DD * DEFINE CLUSTER (NAME(POSIX.DEFKEY.EXAMPLE) - VOLUMES(POSIX1) - ZFS CYL(25 0) SHAREOPTIONS(3) - KEYLABEL(PROTKEY.AES.SECURE.KEY.32BYTE)) /* 15 15

17 zfs Encryption zfsadm define and assign example zfsadm define -aggr POSIX.DEFKEYL cyl 3 -keylabel PROTKEY.AES.SECURE.KEY.32BYTE 16 16

18 zfs Encryption New Encrypted filesystem: FORMAT New keyword(s) on format: (zfsadm format, IOEFSUTL format, IOEAGFMT) -encrypt -noencrypt New zfs configuration variable introduced in V2R3 to set the installation default for encryption (IOEFSPRM): FORMAT_ENCRYPTION ON OFF Specifying encrypt or noencrypt on define will override FORMAT_ENCRYPTION setting 17 17

19 zfs Encryption zfsadm format example zfsadm format -aggr POSIX.DEFKEYL encrypt Must be a version 1.5 aggregate! In V2R3 format_version default is 5, so I did not need to specify the version in this example 18 18

20 zfs Encryption IOEFSUTL format example //ZFMTKEY JOB,'FSUTL V5 KEYL', // CLASS=A,MSGCLASS=X,MSGLEVEL=(1,1),NOTIFY=&SYSUID //* //UTLFMTK EXEC PGM=IOEFSUTL,REGION=0M, // PARM=('format -aggregate POSIX.DEFKEY.EXAMPLE -encrypt') //IOEFSPRM DD DSN=CFCIMGKA.PARMLIB(PARMFILE),DISP=SHR //SYSPRINT DD SYSOUT=H //STDOUT DD SYSOUT=H //STDERR DD SYSOUT=H //SYSUDUMP DD SYSOUT=H //CEEDUMP DD SYSOUT=H //* 19 19

21 zfs Encryption IOEAGFMT format example //ZFMTKEY JOB,'AGFMT V5 KEYL', // CLASS=A,MSGCLASS=X,MSGLEVEL=(1,1),NOTIFY=&SYSUID //* //AGFMTK EXEC PGM=IOEAGFMT,REGION=0M, // PARM=('-aggregate POSIX.DEFKEY.EXAMPLE -encrypt') //IOEFSPRM DD DSN=CFCIMGKA.PARMLIB(PARMFILE),DISP=SHR //SYSPRINT DD SYSOUT=H //STDOUT DD SYSOUT=H //STDERR DD SYSOUT=H //SYSUDUMP DD SYSOUT=H //CEEDUMP DD SYSOUT=H //* 20 20

22 zfs Encryption Encrypting an EXISTING filesystem Assign a keylabel and encrypt with zfsadm encrypt: zfsadm encrypt aggr POSIX.EXIST.FILESYS keyl PROTKEY.AES.SECURE.KEY.32BYTE The existing filesystem must be mounted in read/write mode The filesystem does not need to be SMS-managed or extended format. IMPORTANT: before an existing filesystem is encrypted for the first time, do a full backup of the filesystem! 21 21

23 zfs Encryption querying encryption status zfsadm fileinfo: indicates if a file is encrypted or not zfsadm fsinfo indicates filesystem status: encrypted in-progress % encrypted, or if it has been stopped timestamp when long running command was started & task ID not encrypted 22 22

24 zfs Encryption May be cancelled: filesystem will remain partially encrypted Performance will be degraded for partially encrypted filesystem don t let filesystem remain in this state! Encryption may be resumed (using zfsadm encrypt command) 23 23

25 zfs Encryption fileinfo: encryption status on a file zfsadm fileinfo /ict/mtpt/dir_7/file1 path: /ict/mtpt/dir_7/file1 *** global data *** fid 3313,1 anode 2380,4548 length 6 format INLINE... mtime Aug 4 20:49: atime Aug 4 20:49: ctime Aug 4 20:49: create time Aug 4 20:49: reftime none encrypted not compressed 24 24

26 zfs Encryption Use fsinfo to display encryption status # zfsadm fsinfo -aggr POSIX.MY.FILESYS File System Name: POSIX.MY.FILESYS *** owner information *** Owner: DCEIMGKA Converttov5: OFF,n/a... Status: RW,RS,EN,NC... Legend: RW=Read-write, RS=Mounted RWSHARE, EN=Encrypted, NC=Not compressed 25 25

27 zfs Encryption fsinfo shows progress for encryption in-progress zfsadm fsinfo -aggregate PLEX.DCEIMGNJ.BIGENC File System Name: PLEX.DCEIMGNJ.BIGENC *** owner information ***... Status: RW,RS,EI,NC Encrypt Progress: running, 23% complete started at Nov 21 14:54: task 57F5E0... Legend: RW=Read-write, RS=Mounted RWSHARE, EI=Partially Encrypted NC=Not compressed 26 26

28 zfs Encryption Cancel an in-progress encryption: zfsadm encrypt -aggregate PLEX.DCEIMGNJ.BIGENC -cancel IOEZ00892I Aggregate PLEX.DCEIMGNJ.BIGENC encrypt or decrypt successfully canceled

29 zfs Encryption fsinfo shows progress for stopped encryption zfsadm fsinfo -aggregate PLEX.DCEIMGNJ.BIGENC File System Name: PLEX.DCEIMGNJ.BIGENC *** owner information ***... Status: RW,RS,EI,NC Encrypt Progress: stopped, 23%... Legend: RW=Read-write, RS=Mounted RWSHARE, EI=Partially encrypted NC=Not compressed 28 28

30 zfs Encryption To decrypt zfsadm decrypt -aggr POSIX.MY.FILESYS IOEZ00878I Aggregate POSIX.MY.FILESYS is successfully decrypted. Should be a rare event only employ encryption at your installation when fully ready 29 29

31 zfs Encryption Encrypt / Decrypt authorization requirements: Superuser Authorization Required: Logged in as a uid-0 user -or- Have READ authority to the SUPERUSER.FILESYS.PFSCTL resource in the UNIXPRIV class. Batch formatting with encryption (IOEAGFMT / IOEFSUTL format) also requires READ access to key label 30 30

32 zfs Compression Compress the contents of a filesystem. (i.e., frees up space) filesystem size remains the same. Average amount of disk space that is saved per file is approximately 65% dependent on the type of data that is being compressed

33 zfs Compression Any file larger than 8K can be compressed zfs compresses file data only, not directories Compression is advisory If the compression of a file does not reduce space utilization, the file is left in its uncompressed format. Always eligible for compression if file changes and space will be saved, the file is compressed zfsadm fileinfo - will indicate how many blocks were saved for compression for a particular file

34 zfs Compression Utilizes zedc Authorized Services (introduced in z/os V2R1 as a new priced feature, and needs the supporting HW features) Uses (new to V2R3) configuration option: EDC_BUFFER_POOL Specifies size of real storage reserved for encryption & compression I/O Default: 32 M for zfs PFS, setting may range from 1 M to 1 G edcfixed - IBM Recommended new option for user_cache_size parameter. Allows zfs to permanently fix the user cache file to eliminate data movement from/to auxiliary buffer pool user_cache_size=64m,edcfixed 33 33

35 zfs Creating NEW compression eligible filesystems 1. Define as usual 2. Format for compression New zfs configuration variable FORMAT_COMPRESSION = ON OFF May be overwritten by using compress -nocompress on all flavors of format 34 34

36 zfs Compression zfsadm format example zfsadm format -aggr POSIX.NEWFYS.COMPRESS compress Note: in V2R3 version5 is the default version for format 35 35

37 zfs Compression IOEFSUTL format example //ZDEFFMT JOB,'DEFINE AND FORMAT', // MSGCLASS=H, // CLASS=A, // TIME=(1440),MSGLEVEL=(1,1) //* //* DEFINE FORMAT COMPRESSION ELIGIBLE //* //DEFINE EXEC PGM=IDCAMS //SYSPRINT DD SYSOUT=H //SYSUDUMP DD SYSOUT=H //AMSDUMP DD SYSOUT=H //DASD0 DD DISP=OLD,UNIT=3390,VOL=SER=POSIX1 //SYSIN DD * DEFINE CLUSTER (NAME(POSIX.NEWFS.COMPRESS) - VOLUMES(POSIX1) - ZFS CYL(25 0) SHAREOPTIONS(3)) /* //CREATE EXEC PGM=IOEFSUTL,REGION=0M, // PARM=('format -aggregate POSIX.NEWFS.COMPRESS -compress') //SYSPRINT DD SYSOUT=H //STDOUT DD SYSOUT=H //STDERR DD SYSOUT=H //SYSUDUMP DD SYSOUT=H //CEEDUMP DD SYSOUT=H 36 36

38 zfs Compression Compressing EXISTING filesystems zfsadm compress aggr POSIX.EXIST.FILESYS The filesystem containing the data to be compressed must be mounted read / write IMPORTANT: before an existing filesystem is compressed for the first time, do a full backup of the filesystem! 37 37

39 zfs Compression querying compression status zfsadm fileinfo: indicates if a file is compresed & and space savings compression eligible not compressed zfsadm fsinfo indicates filesystem status: compressed in-progress % compressed, or if it has been stopped timestamp when long running command was started & task ID not compressed 38 38

40 zfs Compression FILEINFO shows compression status of a file This example shows a file that had space savings due to compression zfsadm fileinfo path testmtpt/file4 path: /home/suimgju/c81500/testmtpt/file4 *** global data ***... Mtime Jan 19 12:27: atime Jan 19 12:27: ctime Jan 19 12:27: create time Jan 19 12:27: reftime none not encrypted compressed 24K saved 39 39

41 zfs Compression FILEINFO shows compression status of a file This example shows a file that had no space saving zfsadm fileinfo /ict/mtpt/dir_23/large_file_23 path: /ict/mtpt/dir_23/large_file_23 *** global data *** fid 1536,1 anode 6668, mtime Jul 8 12:35: atime Jul 24 19:27: ctime Jul 27 15:46: create time Jul 27 15:46: reftime none not encrypted compress-eligible 0K saved 40 40

42 zfs Compression FSINFO shows compression status of filesystem # zfsadm fsinfo -aggr POSIX.MY.FILESYS File System Name: POSIX.MY.FILESYS *** owner information *** Owner: DCEIMGKA Converttov5: OFF,n/a... Status: RW,RS,NE,CO... Legend: RW=Read-write, RS=Mounted RWSHARE, NE=Not encrypted, CO=Compressed 41 41

43 Compression FSINFO showing compression in-progress zfsadm fsinfo -aggregate PLEX.DCEIMGNJ.BIGENC File System Name: PLEX.DCEIMGNJ.BIGENC *** owner information ***.. Status: RW,RS,NE,CI.. Compress Progress: running, 48% started at Nov 21 16:34: task 57F5E0. Legend: RW=Read-write, RS=Mounted RWSHARE, NE=Not encrypted CI=Partially compressed 42 42

44 zfs Compression FSINFO showing compression stopped (by using a cancel option) zfsadm fsinfo -aggregate PLEX.DCEIMGNJ.BIGENC File System Name: PLEX.DCEIMGNJ.BIGENC *** owner information ***.. Status: RW,RS,NE,CI.. Compress Progress: stopped, 48%. Legend: RW=Read-write, RS=Mounted RWSHARE, NE=Not encrypted CI=Partially compressed 43 43

45 zfs Compression ZFS_VERIFY_COMPRESSION_HEALTH New health check added for V2R3 compression support Checks whether all user cache pages are registered with the zedc Express service (hardware feature) when there are compressed filesystems. For best performance, all systems accessing the compressed data should have zedc Express installed

46 zfs Compression MODIFY ZFS,QUERY & zfsadm query Help to determine the overall compression effectiveness and allows monitoring of the performance of the zedc. Compress report is included with the ALL & LFS reports # of compression / decompression calls Avg call time KB Input : # of kilobytes sent from zedc card for compression or decompression KB output: # of kilobytes received from zedc card for compression or decompression 45 45

47 zfs Decompression zfsadm decompress aggregate name [-cancel] Should be a rare event only employ compression at your installation when fully ready 46 46

48 zfs Compress / Decompress authorization requirements: Superuser Authorization Required: Logged in as a uid-0 user -or- Have READ authority to the SUPERUSER.FILESYS.PFSCTL resource in the UNIXPRIV FACILITY class

49 zfs Shrink Reduces the physical size of the file system. Releases a specified amount of free space from the VSAM linear data set Filesystem must be mounted read/write Long running command: FSINFO shows progress 48 48

50 zfs Shrink zfsadm shrink aggregate name { size Kbytes [-noai] -cancel} [-trace file_name] [-level] [-help] where kbytes: new total size of the shrunken filesystem -noai means no active increase Active increase is the default behavior for shrink 49 49

51 zfs Shrink: Active Increase Since applications can access the filesystem for most of the shrink processing, there may be additional blocks to be allocated. If this allocation causes more space to be needed in the aggregate that the new total size specified in size, zfs will actively increase the new total size. The shrink command ends in error if the size is actively increased to the original size of the file system. If noai ( no active increase ) is specified on the shrink call and an active increase is needed, the shrink command ends in error

52 zfs Shrink Processing Moves blocks that are in the portion of the dataset to be released into the portion that remains. Requires scanning each internal filesystem structure to determine if it owns any blocks that need to be moved. The bulk of the time in a shrink operation occurs during this scan After block movement completes, the free space is released. Filesystem is quiesced briefly while free space is released 51 51

53 zfs Shrink Use fsinfo output to estimate a reasonable size for shrink zfsadm fsinfo will show: Aggregate size (K) Free size (8K blocks) 1K fragments Don t make -size so small that aggregate will need to grow when regular activity occurs! The difference between the new total size and the current size of the filesystem cannot be larger than the free size 52 52

54 zfs Shrink Example: Based on zfsadm fsinfo results, reasonable shrink size is 500,000 K bytes zfsadm shrink -aggr ZFSAGGR.BIGZFS.SHRINK.EXAMPLE -size

55 zfs Shrink example (post shrink) 54 54

56 zfs Shrink: cancel zfsadm shrink aggr <aggrname> -cancel If interrupted (via shutdown, unmount force, zfs owner goes down) A new zfs owner after interruption will not continue the shrink operation 55 55

57 zfs Shrink authorization requirements: UPDATE (or higher) authority to the VSAM LDS For a filesystem defined with guaranteed space, the ability to shrink will also be controlled by the management class conditional partial release

58 zfs Shrink Rules DFSMS Management Classes If no management class for an aggregate, shrink allowed. If management class exists for aggregate If aggregate defined with guaranteed space, the conditional partial release setting in the management class determines if shrink is allowed. If aggregate not defined with guaranteed space, the partial release setting in the management class determines if shrink is allowed

59 zfs SMF Support zfs as of V2R3 will now utilize System Management Facility (SMF) accounting capability. zfs V2R3 allows for monitoring of: Important file system events (grow, shrink, etc.) General performance information similar to what is seen in query command output. Per-file system performance information

60 zfs SMF Support Record type 92 will contain the zfs subtypes SMF recording in zfs is enabled via a new config variable smf_recording: Options are OFF ON ON,xx where xx is reporting interval in minutes, range 1-60, default 30 This IOEFSPRM value may also be changed with zfsadm config 59 59

61 zfs SMF 92 Support The following are the record subtypes zfs will use: When event occurs: 50 File system events (grow, shrink, etc.) Interval based: 51 zfs call rates and response times 52 zfs user cache statistics 53 zfs meta cache statistics 54 zfs locking statistics 55 zfs IO statistics 56 zfs token manager statistics 57 zfs storage usage summary 58 zfs sysplex transmit statistics 59 zfs per-file system statistics, may be too often 60 60

62 zfs SMF Support zfs will record records in two ways: Interval based records, subtypes Cut at an interval provided by the user via config option smf_recording or by SMF s interval value if none given. File system event subtype 50. Cut when the event occurs. Most of the interval based records contain similar information to that obtained by the zfs query commands. Subtype 59 writes out file system specific information and can be very high volume. It is recommended that this subtype NOT be used unless necessary

63 zfs Implicit format on mount Allows mount of an unformatted VSAM LDS Current process: 1. Create the VSAM LDS (via zfsadm define or DEFINE CLUSTER) 2. Format it as a zfs aggregate (zfsadm format / IOEFSUTL format / IOEAGFMT) 3. mount aggregate Process now allowed as of V2R3 1. Create the VSAM LDS zfsadm define DEFINE CLUSTER --- must use new ZFS keyword 2. mount aggregate Ensure that the catalog is writable, for where the zfs is to be cataloged (most likely a user catalog)

64 zfs Implicit format on mount create zfs 63 63

65 zfs Implicit format on mount Uses installation default for all format parameters format_aggrversion (4 5) new default on V2R3 format_compression (on off) new in V2R3 format_encryption (on off) new in V2R3 format_perms default: o775 new in V2R3 Specifies the default permissions for the root directory of the aggregate for all flavors of format (zfsadm format, IOEAGFMT, IOEFSUTL format & implicit format on mount) when perms is not used

66 zfs Online Salvage Only run from a V2R3 system. A new long-running command which may be used by a systems programmer to verify and / or repair a damaged filesystem, while the filesystem is still mounted. Damaged filesystems are extremely rare! Prior to V2R3 salvage was only available as a batch tool (IOEFSUTL or IOEAGSLV), which required the filesystem be unmounted to run. This V2R3 addition will allow a filesystem to be repaired while still mounted in the rare event this is necessary

67 zfs Online Salvage Use only when the aggregate cannot be unmounted for repairs. No writes are allowed to the aggregate while a salvage operation is running Verification portion (longest part) of salvage may be interrupted: Using the cancel keyword Unmount immediate with the force option During a shutdown Once repair has begun, salvage cannot be interrupted

68 zfs Online Salvage - Tip If a user wishes to check a file system for corruption and NOT block write activity they could use a simple command like: find. name * print exec grep e xxxx {} \; Where xxxx is any string NOT found in any file. This would scan all directories and files and if you get no assertions you likely have a good file system. Online salvage should only be used as a safety valve to verify and / or repair a damaged file system that cannot be unmounted

69 zfs Various Other Enhancements Changes in defaults New configuration options zfs authorization changes zfsadm chaggr to change attributes of a mounted file system Miscellaneous other changes, trace 68 68

70 zfs Changes in defaults format_aggrversion: changed from 4 to 5 change_aggrversion_on_mount: changed from OFF to ON honor_syslist: obsolete in V2R3 zfs V2R3 always honors the USS automove options 69 69

71 zfs New configuration options format_perms format_compression format_encryption edc_buffer_pool edc_fixed option added to user_cache_size smf_recording long_cmd_threads 70 70

72 zfs Authorization changes Format authorization requirement has been changed to require UPDATE authority to the VSAM data set To set owner, groupid, or permissions on format, also require UID 0 or have READ authority to SUPERUSER.FILESYS.PFSCTL 71 71

73 zfsadm chaggr New admin command which changes the attributes of a mounted aggregate aggrfull {threshold,increment OFF} aggrgrow {ON OFF} rwshare norwshare Causes a samemode remount (if mounted read write) Prior to this support, to change these attributes the aggregate must be unmounted and re-mounted

74 zfsadm chaggr syntax zfsadm chaggr -aggregate name {-aggrfull{threshold,increment OFF} -aggrgrow {ON OFF} -rwshare -norwshare} [-trace file_name][-level][-help] 73 73

75 zfsadm chaggr Can only be used when all sysplex members are at z/os V2R3 Only one of aggrfull, -aggrgrow, -rwshare, -norwshare can be specified at a time 74 74

76 -aggrfull Specifies the threshold and increment for reporting aggregate utilization messages to the operator. May be changed to: threshold,increment or OFF 75 75

77 -aggrgrow Specifies whether aggregates can be dynamically extended when they become full chaggr may set it ON or OFF Aggregate must have a secondary allocation specified (when created) and there must be space on the volumes 76 76

78 -RWSHARE and -NORWSHARE Both are read / write mounts RWSHARE is sysplex-aware Filesystem is locally mounted on every member of the sysplex File requests are handled locally by zfs NORWSHARE is sysplex-unaware File operations must be function-shipped to OMVS owning system 77 77

79 zfsadm chaggr Also will update the zfs mount parameters in the z/os UNIX couple dataset, preserving the changes made with chaggr if aggregate is remounted (note: all mount parameters are lost if an aggregate is unmounted) 78 78

80 zfsadm chaggr: example (BEFORE chaggr) 79 79

81 zfsadm chaggr: example 1. zfsadm chaggr -aggregate ZFSAGGR.BIGZFS.FS1 -aggrgrow off 2. zfsadm chaggr -aggregate ZFSAGGR.BIGZFS.FS1 -aggrfull 90,5 3. zfsadm chaggr -aggregate ZFSAGGR.BIGZFS.FS1 -norwshare 80 80

82 zfsadm chaggr: example (AFTER chaggr) 81 81

83 zfsadm chaggr: authorization requirements Superuser Authorization Required: Logged in as a uid-0 user -or- Have READ authority to the SUPERUSER.FILESYS.PFSCTL resource in the UNIXPRIV racf class

84 zfs miscellaneous changes zfsadm new option: -trace file_name Specifies the name of a where trace records from zfsadm call will be put z/os Unix file An existing sequential dataset A member of an existing PDS or PDSE An option on ALL zfsadm commands zfsadm commands that query information: no longer need READ access to IOEFSPRM file (if used) Installations using parmlib (IOEPRMxx) never needed any special authorization 83 83

85 zfs V2R3 Performance Official IBM Performance Measures Your mileage may vary Measurements were taken on a IBM zec12 (model 2827) with 4 CPUs 84 84

86 zfs Encryption Performance Measurements 85 85

87 zfs Compression Performance Measurements 86 86

88 zfs Shrink Performance Measurements 87 87

89 zfs Online Salvage Performance Measurements 88 88

90 zfs Overall V2R3 Performance (compared to V2R2) Description of workloads used: FSPT FS DU DL DL2 DUC DL3 file I/O workload. random read/writes (typically 65% reads/35% writes, some with 100% reads) file I/O workload. Mix of random and sequential reads / writes, creates & deletes files Directory reads & writes Directory search & readdir Directory search Intensive directory update workload Directory workload, readdir DL3 plus Directory workload, extended readdir operations 89 89

91 zfs Overall V2R3 Performance (compared to V2R2) 90 90

92 z/os UNIX Enhancements HFS to zfs utility Automatic unmount of version file system in shared file system Maintaining a single configuration file across the sysplex Assigning default umask values to all z/os UNIX users 91 91

93 z/os UNIX HFS to zfs utility HFS has been stabilized for quite some time. Support for HFS is planned to be removed in It is therefore necessary for data residing in HFS file systems to migrate to zfs file systems. Tools such as BPXWH2Z have been provided, however, file system must effectively be taken offline to perform the migration. This approach has not been acceptable for data for some critical applications. An online migration facility is provided to replicate and replace an HFS file system with a zfs file system, while mounted and in use. Use of this facility obviates the need to take application or system outages to migrate data off of HFS Availability is with the PTF for APAR OA

94 z/os UNIX HFS to zfs utility Command to initiate, modify, and monitor migration is BPXWMIGF BPXWMIGF is available from TSO, the UNIX shell, and console (SYSREXX) Migration occurs in two phases: 1. Data is mirrored and the active mirror is maintained until swap. Applications continue to run and are unaffected during this phase 2. The zfs file system is swapped in replacing the HFS Access to the file system is briefly quiesced during the swap These phases can be scheduled independently Migration must be initiated from the HFS owning system BPXWMIGF utility has three basic functions 1. Initiate and modify migration (requires superuser or BPX.SUPERUSER) 2. Query migration 3. Cancel migration (requires superuser or BPX.SUPERUSER) 93 93

95 z/os UNIX HFS to zfs utility 94 94

96 z/os UNIX HFS to zfs utility 95 95

97 z/os UNIX HFS to zfs utility Shell command examples: Start a migration bpxwmigf -source omvs.wjs.hfs -target omvs.wjs.temp -srename omvs.wjs.old -trename omvs.wjs.zfs Check migration status bpxwmigf -query After mirroring complete, update mount statements and do the swap bpxwmigf -source omvs.wjs.hfs -swap Note: you might want to use a temp name for the new zfs just to make sure it does not accidentally get mounted if the migration does not complete 96 96

98 z/os UNIX HFS to zfs utility HFS to zfs requirements, before using: All systems in the OMVS sharing group must be V2.3 Review mount statements or policies for mounting a file system to be migrated HFS file system must be mounted zfs file system must be allocated and formatted, ensure file system is properly sized. Do not mount the zfs file system. (((Ensure that the z/os V2.2 enhancement of zfs in the OMVS address space is not in use.))) Requirement removed with OA53738! Limit migrations to only one at a time until swap. Requirement added with OA After migration: Ensure mount statements or policies will mount the zfs file system If you mount the old HFS, you will have regressed your data

99 z/os UNIX Automatic unmount of version file system in shared file system Once you have rolled off a version file system (new name for the version root file system ), then it was additional manual work to unmount it and everything underneath it. Customer Requirement to unmount a version file system when it is not in use anywhere in the shared file system environment. Specify in BPXPRMxx: VERSION('nnnn'[,UNMOUNT NOUNMOUNT]) VERSION statement applies only to systems that are exploiting shared file systems. UNMOUNT to turn on the version auto-unmount option, which will cause the VERSION_AUTO_UNMOUNT attribute to be set for the version file system if it is mounted on this system. You can also dynamically turn on or off the VERSION_AUTO_UNMOUNT attribute using: SETOMVS VERSION=('string',UNMOUNT NOUNMOUNT) You D can MVS,[,VERSION=(*,UNMOUNT NOUNMOUNT]) also change the settings of [,VERSION=(*,UNMOUNT NOUNMOUNT]) this parameter by using SET OMVS=(xx) and SETOMVS RESET=(xx) parmlib specifications

100 z/os UNIX Maintaining a single configuration file across the sysplex New directory, /global, is available in the sysplex root, for all systems: can be used to store and maintain a single copy of a configuration file that can be referenced by all members of the sysplex. Example: Need to maintain a consistent automount policy for use by the automount daemon on every member of the sysplex (/global/auto.master and /global/u.map for the /u directory) For convenience, provide a view of program products at multiple levels, even though the product may be installed on only one member of the sysplex. Example: Want to provide a simple and consistent way to access multiple levels of Java (Java 6, Java 7, Java 8) or COBOL. (/global/java/java8 > symlink --> /zos23/usr/lpp/java8) Some users already might have done something similar. Prior to V2R3, individual copies of the same configuration file had to be maintained in each member of the sysplex. /global will have no DDDEFs and will not be pointed to for any default installation. (For consistency, there is a /global added to the version file system also.) 99 99

101 z/os UNIX Assign default umask values to all z/os UNIX users New UMASK statement of the BPXPRMxx parmlib member can be used to assign a default umask value for all z/os UNIX users. UMASK(nnnn) nnnn is a 3-to-4 digit octal number, or NONE, that specifies the permission settings of files or directories created by users or processes that are not to be allowed. If a umask value is not specified in BPXPRMxx, the default is NONE, which indicates that no umask default is in effect. See the Value Range table in MVS Init and Tuning, for instance, UMASK(007) will give no access to Other. Use this setting to better control the files and directories that users on your system can access. Existing facilities (/bin/sh for instance) that set the umask value will continue to override the system default. Changing the default umask value might have unexpected results on your applications. A umask value that is consistent with the umask value that is set in /etc/profile might be appropriate. Rolled back to z/os V2R2, with APAR OA51000 installed

102 Summary zfs: Encryption and Compression only when on z/os V2.3 Shrink free up some space! SMF Recording data galore Implicit Format on mount one less thing to worry about Online Salvage (verify / repair) for when you really need it Supplemental: Dynamic Change Aggregate Attributes (chaggr), Miscellaneous updates, and zfs V2R3 Performance Measurements z/os UNIX: HFS to zfs utility intended for those file systems that cannot be unmounted Automatic unmount of version file system in shared file system let the system manage and monitor this Maintaining a single configuration file across the sysplex convenience Assigning default umask values to all z/os UNIX users handy, but considerations are necessary

103 Appendix zfs Publication references z/os Distributed File Service zseries File System Administration z/os Distributed File Service Messages and Codes z/os DFSMSdfp Storage Administration z/os UNIX System Services Command Reference z/os MVS System Management Facilities (SMF) z/os UNIX Publication references z/os UNIX System Services Planning z/os UNIX System Services Command Reference z/os Distributed File Service zfs Administration z/os MVS System Messages, Vol 3 (ASB-BPX) z/os UNIX System Services Messages and Codes 102