Skip to content

Latest commit

 

History

History
61 lines (46 loc) · 2.73 KB

aix_ssp.md

File metadata and controls

61 lines (46 loc) · 2.73 KB
Raw

Shared Storage Pools (SSP)

VIO Command Description
cluster -list show cluster name (a good way to see if VIO is part of cluster)
cluster -status -clustername mycluster -verbose more detail

  • ON_LEVEL means pool at same level of VIO software.

  • UP_LEVEL means software is higher than pool. Once all VIOs go UP_LEVEL they will negotiate and pool will go up to new level, with new features.

VIO Command Description
lscluster -d Lots of info. Shows names of disks on all the VIO servers for storage pool. CLUSDISK is a cluster disk, i.e. a normal disk in the pool (0.25TB luns might be typical), REPDISK is a repository disk, which is how the VIOs talk to each other.
pv shows pv command usage
pv -list shows disks.
pv -list -capable shows what luns are available on all
failgrp -list Show mirror groups. FG_NAME of Default means that it's not mirrored.
failgrp -create -fg myfgname: hdisk12 hdisk13 Checks that luns are available on all VIOs, adds them to pool and starts mirroring.
failgrp -modify -fg Default -attr fg_name=new_name rename fg
failgrp -list
failgrp -list -verbose
pv -list show which fgs the disks are in.
pv -add -fg v7000tan: hdisk10 hdisk11 v7000dun: hdisk14 hdisk15 Adds luns then starts spreading disk blocks around.
pv -list -verbose shows more info, e.g.. MPIO 2076 FC Disk
lsmap -all will show name of lu as backing device.
lu -list shows lu names (typically use client lpar names)
lu -create -lu orange4a -size 64G -vadapter vhost3 creates a new lu as a backing device (very quick) and maps it to client (vhost).
lu -create -lu orange4a -size 111G -vadapter vhost3 creates another with same name. Beware.
lu -remove -lu orange4a error because not unique.
lu -remove -luudid blahdeblahdeblah need long id number if name not unique.

Thin Provisioning

Default provisioning is thin. Can specify thick on lu -create command line.

Snapshots

Can stop an lpar, take a disk snapshot & then bring it back up.

Repository Disk

Can use <1GB disk. Uses tiny part at beginning of disk. VIOs only use repo disk when changes are made so not used all the time.. One VIO at a time is designated as the one which can write to the repo disk.

chrepos to move from one disk to another.

If the repository disk breaks, just use chrepos. It will see that old disk is bad and write new stuff on the new disk.

If a VIO is down at the time, it will negotiate with other VIOs and work out what went wrong. No need for intervention.

In any event, give it time to sort itself out.

SSP will continue to work without repo disk, but good to monitor error logs and have a spare repo disk on hand just in case.