Log Shipping + Database Snapshots == Bummer, dude!!

reedme — Fri, 24 Apr 2009 18:29:47 GMT

So, like most funny stories (involving me), this one starts with “Yesterday, I had this bright idea...” and goes downhill from there.

Database mirroring is great. Full stop. But there’s this one little catch: it can only have a single mirror copy of the principal. Awesome for HA, but not so good for scale out.

Because database mirroring with SQL Server 2005 and 2008 makes available this great way to leverage your failover partner database for reporting purposes (to get some value out of your high availability solution besides insurance and a good night’s sleep) using database snapshots... well, you just had the same idea that I did, right?

It only seemed logical that log shipping would allow for the same AND allow for multiple target servers/databases, right?

Not so much.

With 2005, the log shipping target in a warm standby mode is in a perpetually restoring state (sort of like a mirror copy, eh? *grumble*), but you can’t take a snap of a database in restoring mode. Of course, I tried it. Wouldn’t you?

And even though Katmai exposes the feature in the UI now that allows for a log shipping target database to be online and query-able in read-only mode (if memory serves, you could do it the “hard way” in Yukon), it’s not possible to take a database snap shot of it... because it’s read-only. Did you know that creating a database snapshot actually modifies the database being snapped? I’d never thought about it. And now I know the answer!

The error message when creating the snap of a read-only database isn’t intuitive, either (if you read from the bottom up – the first one’s pretty clear – don’t ask): “A database snapshot cannot be created because it failed to start.” Huh?

Read-only standby mode shows some promised BUT you’ve got to close all open connections to the database and put it in a restoring mode in order to restore future tran log backups to it as part of your log shipping. Doh!

Needless to say, it would be pretty handy [if this worked] for a scale out scenario where I need to keep, oh, say, 44 copies of the same database continuously in sync AND available 100% of the time for customers to query, but nooo... It couldn’t possibly be that simple.

But you know, I probably wouldn’t be happy (or have a fun job) if it was that simple. Heh. Besides, not I get to go Visio up a more complicated solution, right?

January is Windows Server 2008 High Availability Month

reedme — Tue, 08 Jan 2008 23:31:31 GMT

If you've been meaning to ramp up on Windows Server 2008 clustering, but haven't made the lab time yet, here's the easy way to do it...

Virtual Lab

Windows Server 2008 Enterprise Failover Clustering

January Webcasts

TechNet Webcast: Building High Availability Infrastructures with Windows Server 2008 Failover Clustering
Monday January 21, 2008, 11:30 AM – 1:00 PM Pacific Time

TechNet Webcast: Failover Clustering 101
Tuesday January 22, 2008, 11:30 AM – 1:00 PM Pacific Time

TechNet Webcast: Failover Cluster Validation and Troubleshooting with Windows Server 2008
Thursday January 24, 2008, 9:30 AM – 11:00 AM Pacific Time

TechNet Webcast: Geographically Dispersed Failover Clustering in Windows Server 2008 Enterprise
Monday January 28, 2008, 11:30 AM – 1:00 PM Pacific Time

TechNet Webcast: High Availability with Hyper-V
Wednesday January 30, 2008, 9:30 AM – 11:00 AM Pacific Time

TechNet Webcast: Deep Dive on Failover Clustering in Windows Server 2008 Enterprise Storage and Understanding Quorum
Thursday January 31, 2008, 9:30 AM – 11:00 AM Pacific Time

The best part? The price.

Reed Me : high availability

Log Shipping + Database Snapshots == Bummer, dude!!

January is Windows Server 2008 High Availability Month