Comments on: Problems We Solve

By: Brent Ozar

Brent Ozar — Tue, 09 Oct 2012 09:29:27 +0000

Rohan – we worked with Microsoft to recreate the scenario, and they had difficulties with it, so I don’t think it’s productive for us to try here. We spent weeks working with them, and explaining the whole scenario is beyond what I can do in a blog comment. It’s pretty elaborate. Thanks though!

By: Rohan

Rohan — Tue, 09 Oct 2012 07:03:41 +0000

On the worker threads issue, were the HADRthreads in SUSPEND, if so what caused them to go into suspension?

So are you saying that once there were available worker, the threads responsible for syncing AG data did not come out of suspension, and did so only after the AG was reset because of change in settings?

Was the Secondary Sync or Async?

How much time did you wait before Nick changed the settings?

I ask all this because I am trying to recreate the same issue with AG for similar loads.

By: Jerry Cutshaw

Jerry Cutshaw — Tue, 02 Oct 2012 13:48:33 +0000

That thought occurred to me as well. For short term recoverability I’ll be snapshotting the VMs. For long term recoverability I’ll leverage vRanger backups (nightly) to cover the OS/Server backups and of course a good SQL Server backup policy.

The only bozo that would fill up the C drive is is me! I’m told that with Windows 2008R2 you can now extend any volume (even the C volume through diskpart).

Thanks for the response. Much love.

By: Brent Ozar

Brent Ozar — Tue, 02 Oct 2012 13:19:54 +0000

Jerry – how do you plan to handle Windows patches, SQL patches, and full-C-drive conditions? (Like some bozo downloads a bunch of ISO files to the C drive or Windows Updates fill it up and knock the machine offline.)

By: Jerry Cutshaw

Jerry Cutshaw — Tue, 02 Oct 2012 13:17:51 +0000

We are re-architecting our SQL infrastructure and this time around we’ve decided to go with stand-alone SQL Servers that take full advantage of VMWare clustering for HA. The reasoning. . .fewer complexities related to clustering setup/maintenance, immediate failover isn’t a necessity and we don’t have the necessity to scale-up vs. scale-out so we can take advantage of commodity hardware (dual hex-core w/ 96 GB RAM).

Does anyone familiar with these variables have any input for or against this solution?

By: Paul Brewer

Paul Brewer — Thu, 27 Sep 2012 14:58:39 +0000

Image trying to implement and support it on servers using Windows Core.

By: Brian G

Brian G — Tue, 25 Sep 2012 13:05:12 +0000

Brent – Thanks for sharing your experiences with the overwhelming majority of us for whom AlwaysOn Availability Groups remain something we are only familiar with on an academic level. A very generous contribution by you. It amazes me how many times I have to learn that there is no easy button for these types of complex problems. MS marketing fools you once, shame on them. MS marketing fools you every three years for your entire career… well…

By: Subhash Pant

Subhash Pant — Tue, 25 Sep 2012 03:06:43 +0000

Thank you for posting this. Though we are not in 2012 yet, always a pleasure to expand knowledge and learn from a guy who has made expensive mistakes on other people’s hardware. Hehe

By: Jason Horner

Jason Horner — Mon, 24 Sep 2012 19:16:54 +0000

ahh ok, I thought maybe it was due to the MTBF characteristics of the two media types. I was reading too deep into the diagram. Thanks for clarifying.

By: Sreekanth

Sreekanth — Mon, 24 Sep 2012 18:48:20 +0000

Thanks Brent!