Download presentation
Presentation is loading. Please wait.
Published byJared Pearson Modified over 9 years ago
1
Storage Management: MirrorDisk/UX, Online JFS Introduction LVM Striping Mirror Disk / UX – Basics Online Backups (lvsplit) Mirror Disk / UX – Advanced Mirrored Boot Devices Snapshots
2
Introducing the Journal File System Base Version Bundled Veritas Extended File System (VXFS) Fast fsck Extent-based allocation Enhanced mount options On-line JFS Additional Features On-Line Defragmentation On-Line Backup – Snapshot On-Line Growth/Reduction
3
Software Disk Striping When you use disk striping you create a Logical Volume that spans multiple disks, allowing successive blocks of logical extents to map to different disks. Valid stripe sizes are 4k, 8k. 16k, 32k, and 64k. The default size is 16k. You should match the stripe size with the block size of the file system (or a number evenly divisible by the stripe size). EG: If the planned block size for a filesystem is 16k, the stripe size should be 4, 8, or 16k. Note: You cannot use striping and mirroring on the same logical volume.
4
Striping – Benefits and Costs Benefits Data access performed over multiple disks simultaneously I/O can be balanced across interfaces Costs Hardware failure on a single disk may result in loss of portions of many files Entire files must be restored, even though only a portion was lost Added complexity Maximizing the Benefits of Striping
5
To maximize the benefits of striping it is best to spread the volumes across separate interfaces. Best performance results from a striped logical volume that spans similar disks Select the appropriate stripe size For HFS file systems match the block size. For JFS file systems, always use 64K stripes. For RAW data choose a size that matches the primary I/O size for the application. Creating Striped Logical Volumes
6
lvcreate –i stripes –I stripe_size [-l | -L space] [-n name] VG -inumber of disks to stripe across -Istripe size in kilobytes (4, 8, 16, 32, or 64) -ltotal space expressed in extents -Ltotal space expressed in megabytes -nname of the logical volume EG:lvcreate –i 3 –I 8 –L 24 –n lvstripe /dev/vg01
7
Implementing Mirroring
8
Mirrored I/O Scheduling
9
Protecting from Disk Failure PV0 Scheduled - returned PV1 Scheduled - returned PV2 Scheduled, returned – retry – unavailable High priority write of PV status to VGRA on all available pvs.
10
Mirrored Write Cache Tracking of pending I/O is done in a cache called Mirrored Write Cache (MWC). The MWC holds the status of write I/Os to a relatively large areas of a logical volume called Logical Track Groups (LTG). When a write I/O is scheduled it is resolved onto the LTG it affects. This entry in the cache is marked dirty. The I/O is scheduled, and when it completes the I/O is marked clean. The MWC can be turned off Provides a runtime performance gain Mirror resync takes longer At volume group activation if MWC is off then /usr/sbin/nomwcsyncd is started to do a background mirror resync You can never run this manually, if necessary deactivate and reactivate the volume group.
11
LVM Commands for Mirroring lvcreate [–m copies] [-L size in MB] VG lvextend [-m copies] LV PV lvreduce [-m copies] LV PV lvdisplay –v LV You may have 0, 1, or 2 copies of any given logical volume
12
LVM Commands for Synchronization lvsync LV vgsync VG Resynchronizes from one copy to make all mirror copies consistent Automatically done when a volume group that is not currently active, is activated. Rarely done manually.
13
Online Backups In this example you see lvol2 is mirrored with one copy on PV0 and one on PV1. A lvsplit operation will freeze the mirror copy and create a ‘backup’ logical volume which points to it. This can now be mounted and used while activity continues on the other PV. If you have three mirrors when you do this, you can take the backup while continuing to have data protection for the active volume. Once a backup has completed you can use the lvmerge command to resynchronize the mirrors.
14
Base JFS / Online JFS Base JFSOnline JFS Online Backupnot possibleonline, snapshot Extending fsoffline, extendfsonline, fsadm Reducing fsnot possibleonline, fsadm Defragmenting fsnot possibleonline, fsadm Fast recovery fsckyesyes
15
fsadm fsadm –F vxfs [-d] [-D] [-e] [-E] /mountpoint_dir fsadm-options: -F vxfsspecifies the journaled filesystem -dPerforms a directory defragmentation -DReport on directory fragmentation -ePerforms a file extent defragmentation -EReport on file extent fragmentation Regains lost performance Should be periodically used (between once a day and once a month) Can be done manually or scheduled with a cron-job The optimization runs online on a mounted JFS file system
16
Resizing
17
Increasing an Online JFS filesystem Extend the logical volume to the needed size: # lvextend –L new_size lvol_name The value of new_size is given in megabytes! Extend the JFS filesystem online: # fsadm –F vxfs –b new_size /mountpoint The value of new_size is given in kilobytes! Check the size of the increased file system: # bdf
18
Reducing an Online JFS filesystem Reorganize directories/file extents: # fsadm –F vxfs –d –D –e –E /mountpoint Reduce the JFS file system online: # fsadm –F vxfs –b new_size /mountpoint The value of new_size is given in kilobytes! Reduce the logical volume to the needed size: # lvreduce –L new_size lvolname The value of new_size is given in megabytes! Verify the size of the reduced file system: # bdf
19
Online Backup via Snapshot
20
Creating a snapshot Create a snapshot # mount –F vxfs –o snapof=/orig_vol /aux_vol /aux_dir /orig_volfile system, which you want to take a snapshot of /aux_vollogical volume to hold the snapshot (no file system) /aux_dirdirectory to mount the snapshot Backup from the snapshot file system # cd /aux_dir # tar cvf /dev/rmt/0mor # find. | cpio –ovcx > /dev/rmt/0mor # vxdump –f /dev/rmt/0m /aux_dir Unmount the snapshot, then lvremove the volume
Similar presentations
© 2024 SlidePlayer.com Inc.
All rights reserved.