EU clears Oracle/Sun merger.

http://www.oracle.com/technology/software/products/database/oracle11g/112010_sparc_x64soft.html

Just a quick note / reminder that iView support came in with the 3.1 firmware update, so install that update and tune in. Australian PS3 cultists only.

http://www.abc.net.au/corp/pubs/media/s2662012.htm

Solaris 10 update 8 (10/09) has been released. New features documentation here.

Looks like Sidekick users are having bad weather in their Clowd, which serves as a reminder that hosted services aren’t the magical sure-fire solution to avoiding infrastructure management.

We tend to assume that online facilities, especially those provided by large organizations are going to be reliable, trustworthy and better administered than a small-budget local solution. We expect redundant servers, storage, backups, probably even multi-homed networks.

“They’re still looking for a way to recover it, but they’re not giving users a lot of hope”

Normally, a critical server failure has a recovery plan - “tape time”. When everything fails, rebuild the servers, restore from tape and get the show back on the road. This takes time but the data comes back.

For a major carrier to have an outage with no solid recovery plan, which implies no disaster recovery solution such as an off-site replication target is a surprise - but should it be? These sort of hosted services are pitched as being cheap or free, so should user expectations be so high?

If you’re not paying for a level of service, how can one expect that service to hit the reliability level that you know you’d achieve running it in-house?

When we deploy, for example, an in-house mail solution, we will have (generally) as a minimum

1. Multiple MXes, including a secondary on a remote network
2. Regular backups
3. A restoration plan to make use of those backups

But when we outsource our mail to a third-party provider, unless they actually make a sales position of their infrastructure and preparedness, how do we know if they’ll survive a server failure? A faulted storage array with data corruption?

I think the lesson to be learned from this failure isn’t just for Sidekick users - it’s for everyone using hosted services in a manner in which would lead to a problem if they went away.

How much impact would the loss of email, calendaring, contact management and instant messaging have on your day?

Most desktop workloads that don’t require high-end video or large amounts of I/O seem to be natural VDI candidates. General browsing, office applications and the like really do hum along nicely and really would lend themselves to a Clowd (heh) deployment.

Except one.

Except one really popular desktop application in business.

Outlook.

Outlook 2007 seems to be able to rip-snort through more I/O bandwidth than you’d believe. I have Outlook running in a VDI session for work purposes, and wow - as it crunches through multi-gigabyte .OST and .PST files, it really keeps the I/O roaring pretty heavily. I keep those files on a seperate server to the rest of the VDI storage, because it was competing too heavily for resources. A quick run of iostat -xn on *that* host tells me Outlook is sustaining 10Mbyte/sec of I/O for a long time, especially when “checking” after an unclean termination. I get the feeling it’d eat more if there was more network bandwidth for it to eat.

Now, this leaves me pondering a modest 200 seat VDI deployment. After some sort of server interruption, that could be 200 seats logging in at 9am and firing up Outlook - this could seriously impact network and storage performance and leave a lot of users taking long morning coffee breaks while their system grinds to life.

Comments, anyone? I’d like to think this is a solved problem, given the popularity of Outlook on the desktop.

For some reason, after a reboot the Solaris iSCSI target daemon comes up with a blank list of targets even though the shareiscsi property on the zvols are set correctly. To fix, “svcadm restart iscsitgt” and the problem will clear immediately.

http://mexico.purplecow.org/index.php/VDI:_Failure_to_start_VM_with_iSCSI_errors

Oracle have reduced the core multiplier for UltraSPARC T2+ machines - T5240, T5440 and friends - from 0.75 to 0.50.

http://www.oracle.com/corporate/contracts/library/processor-core-factor-table.pdf

A few quick hints and tips for VDI for the working sysadmin:

• Read the normally-skipped “Getting Started” guide.
• Read the Release Notes.
• Install the patches as soon as possible to avoid later pain.
• Either don’t use ZFS on your VBox host or constrain the ARC to a small size (VDI looks at free memory and counts ARC as used).
• Be careful running Explorer on the VDI host - it seemed to hang some of the VMs. (!)
• SMF everything as soon as possible, because you’ll be rebooting a few times while trying to iron out problems.
• The version of SRSS that ships with VDI will dump firmware on your SunRays that appears to break some older LCD support - broken sync rate and 640×480. Fixable by firmware downgrade.
• Don’t use NAT networking, it’ll break in unpredictable, random, intermittent ways.
• Don’t put spaces in your root password, it’ll break the VBox install script.
• Install Brendan’s dtrace toolkit and use it - execsnoop etc will be handy.
• Don’t trust the error messages implicitly, use truss / other logs to confirm.

Suggestions for the Solaris n00b trying to get some joy out of VDI:

• Understand ZFS, iscsi, vbox, srss before trying to drag it together with VDI
• Understand LDAP before you start.
• Use OpenDS rather than DSEE if you’re not in love with LDAP/DSEE.
• While you’re setting things up, some stuff will *break* with little or no obvious error messages. “init 6″ is a lot easier than trying to diagnose it, but it’ll mask some problems.

More notes are in the wiki at http://mexico.purplecow.org which is mirrored to http://www.purplecow.org

WordPerfect was a pretty classic bit of software. So let’s make vim smell like WordPerfect, for a bit of nostalgia:

WordPerfect looks like this (a quick snatch from Wikipedia, yay):

And this screenshot should give you what you need to make vim look quite similar:

Ignore the syntax and folding lines, they’re just left from my usual .vimrc.