Skip to content

Latest commit

 

History

History
73 lines (58 loc) · 2.21 KB

aix.md

File metadata and controls

73 lines (58 loc) · 2.21 KB

AIX

HDLM

  • dlnkmgr view -drv view dlm devices.
  • dlmrmdev deletes dlm devices. Must be done before deleting underlying hdisk devices.
  • dlmpr -a clears all persistent reservations

Uninstall:

installp -u DLManager.mpio.rte
installp -u Hitachi.hacmp.support
installp -u Hitachi.aix.support.rte

Powerpath

  • emcpreg -add some-key to register powerpath software.
  • lsattr -El fscsixx, may want to see dynamic_tracking and fast fail are on
  • lsattr -El device check q depth (recommend 16) for all EMC or Powerpath devices seen in lsdev -Cc disk

Network

  • no -p -o rfc1323=1 allows setting values greater than 4K; without this, send/recv sizes can be set higher, but will only use first 4K.
  • tcpdump puts interface into promiscuous mode

Misc Tips / Tricks

  • lssyscfg -r lpar -m msys --osrefresh force refresh os information on HMC
  • lslock -e frame-name list which HMC has a lock
  • viosvrcmd -m msys -p vio-name -c errlog run command on vio from HMC
  • viosbr dump VIOS config to a file. Use -view to view as a report.
  • print "cvai" | kdb | grep vscsi | grep -v read show vio name and vhost from client lpar
  • LMB size must be same on all systems for LPM (use lshwres with -F mem_region_size to view)
  • NIMBUG env variable - debug nim
  • alog -f file -o to view alog format files (e.g. nimlog)
  • vmo -o nokilluid=uid_num tell aix not to kill processes owned by this user when out of memory
  • rmcdomainstatus -s ctrmc test rmc connection, I = up, O = down.
  • pmcycles show CPU speed

Power8

  • 12 cores per chip
  • 8 SMT threads
  • S824 up to 2TB
  • CAPI adapters <500 instructions (vs 20K) to get to SSD disk
    • appears like slow memory rather than fast storage

E870 / E880

  • combine 795 & 770 & 780
  • 16TB memory
  • 19" rack
  • mandatory midplane, with nodes (1 to 4) above and below
  • no disk (SAS) in nodes; need IO drawer
  • 32 cores per node; 4 node has 128 cores

sugroups

If user A has sugroups = x,y,z then all users in groups x,y,z can su to user A

Shared Ethernet Adapters

Force a primary SEA (currently in auto mode) to fail over to standby...

chdev -l <SEA device> -a ha_mode=standby

Bring it back...

chdev -dev <SEA device> -attr ha_mode=auto