random musings from a twisted mind
Archive for July, 2007
helping hands with puppets
Jul 26th
Will have to have a good look at [http://puppet.reductivelabs.com/ puppet], looks like a re-invention of cfengine, just using ruby instead (which is a good thing). Some of the recipes look really useful, though might have to write a few for solaris zfs/zones etc.
Just a big shame that the trac pages look so very broken within the theme (does kind of put you off)
Zones
Jul 11th
Configure the zone you want, this is just a basic configuration (suited to a name server – that inherits the global zones /var/named)
# zonecfg -z solzone
solzone: No such zone configured
Use 'create' to begin configuring a new zone.
zonecfg:solzone> create
zonecfg:solzone> set zonepath=/u01/zones/solzone
zonecfg:solzone> set autoboot=true
zonecfg:solzone> add net
zonecfg:solzone:net> set address=10.10.10.20/24
zonecfg:solzone:net> set physical=bge0 zonecfg:solzone:net> end
zonecfg:solzone> add fs
zonecfg:solzone:fs> set type=lofs
zonecfg:solzone:fs> set special=/u01/zonedata/solzone
zonecfg:solzone:fs> set dir=/u01
zonecfg:solzone:fs> end
zonecfg:solzone> add inherit-pkg-dir
zonecfg:solzone:inherit-pkg-dir> set dir=/var/named
zonecfg:solzone:inherit-pkg-dir> end
zonecfg:solzone> add attr
zonecfg:solzone:attr> set name=comment
zonecfg:solzone:attr> set type=string
zonecfg:solzone:attr> set value="test solaris zone"
zonecfg:solzone:attr> end
zonecfg:solzone> verify
zonecfg:solzone> commit
zonecfg:solzone> exit
And now install and boot the zone
# zoneadm -z solzone install
# zoneadm -z solzone boot
db2 gotcha on solaris 10
Jul 10th
Hmm, nice little error when trying to start db2 on solaris 10, stating that it was unable to start up the database and i should check my SHM settings!
Arrgh!
prctl reported correct project settings, so tried truss:-
# truss -f db2start
truss: cannot trace set-id or unreadable object file: db2start
Cue alarm bells, proven by trusty “ls”:
-rwsr-xr-x 1 root root 238584 Jul 10 12:28 db2start
Add root to the user.db2inst1 project with a quick projmod -a -U root user.db2inst1 and try again.
blogs.sun.com posts to follow up on
Jul 5th
Lots of interesting posts on sun.com (and lots of pointers to things that aren’t so interesting 
– but really need to follow up on the following:-
http://blogs.sun.com/chrisg/entry/where_are_all_the_log
http://blogs.sun.com/taylor22/entry/configuring_jumbo_frames_on_the
http://blogs.sun.com/damico/entry/the_death_of_usr_ccs
http://blogs.sun.com/ahl/entry/iscsi_dtrace_provider_and_other
http://blogs.sun.com/marks/entry/zfs_delegated_administration
http://blogs.sun.com/avalon/entry/nevada_solaris_10_update_4
crash dump analysis on solaris x86-64
Jul 4th
Found this written by Frank Hofmann, very interestly little history lesson about the x86 chipset. (he also talks about how HOSTID is generated on solaris x86 – which shows how important it is to keep track of your /kernel/misc/sysinfo file for hostid locked software 
observations with solaris scheduling
Jul 2nd
One quite straight-forward thing – which i didn’t really think about until I was playing with multiple zones and the FSS – was that you can actually have 0 shares allocated – which just means you get all the spare cycles. If you have 5 zones, all of which running FSS (and why wouldn’t you be), all configured like the following:-
| ZoneID | zone Name | Shares |
|---|---|---|
| 0 | Global Zone | 50 |
| 1 | prod | 50 |
| 2 | preprod | 25 |
| 3 | test | 0 |
| 4 | dev | 10 |
Pretty good for when you do want to make sure that any non-production zones should never impact production (well aside from memory and network impact).