I remember setting up (and this was best practice then) five-node clusters – three active, two passive – because once you evicted a node there was no guarantee that you could get it back into the cluster before yet-another server would fail.
Good times. Of course, 9 GB disks and 16 MHz P-IIs were all the rage back then… From: [email protected] [mailto:[email protected]] On Behalf Of Kevin Lundy Sent: Thursday, October 27, 2016 12:59 PM To: [email protected] Subject: Re: [NTSysADM] Advice: break a Win2008 R2/SQL 2008 R2 cluster, and rebuild as 2 single node clusters Gee thanks for the nightmare reminders about Wolfpack! Ughh. Almost as bad as running Decnet on NT. On Thu, Oct 27, 2016 at 12:49 PM, Michael B. Smith <[email protected]<mailto:[email protected]>> wrote: That should work fine. It's not scary like it used to be in the early wolfpack days (Wolfpack was the codename for Windows Failover Clustering). This may be obvious, but remember that a single-node cluster provides you zero protection. :-) -----Original Message----- From: [email protected]<mailto:[email protected]> [mailto:[email protected]<mailto:[email protected]>] On Behalf Of Michael Leone Sent: Thursday, October 27, 2016 8:51 AM To: [email protected]<mailto:[email protected]> Subject: [NTSysADM] Advice: break a Win2008 R2/SQL 2008 R2 cluster, and rebuild as 2 single node clusters I could use some advice, as I've done some of these steps before, but not all, and not on a production cluster. We have a Win2008 R2/SQL 2008 R2 cluster, 2 nodes. We want to break the cluster (turn it into a single node cluster), then take the other node; reformat as Win2012 R2/SQL 2012 R2 cluster with just itself as a single node. This will become a new cluster (at least I hope it will have a new cluster name, different from the existing one ...). Once the new one is up and running, they will migrate the DB from the old cluster to the new. Then they want to then do the same to the current node, and then ad it to the new 2012 R2 cluster as it's second node. The (eventual) end result is to migrate from a Win2008 R2/SQL 2008 R2 cluster, to a Win2012R2/SQL2012R2 multi-site cluster (one node here, one node over at the D/R site). Anyway, I have made a multi-site Win2012 R2/SQL 2012 R2 cluster before (once ..), but I wasn't the guy who broke that existing cluster, so it could be rebuilt. My worry is gracefully breaking the cluster, especially since this is our production SQL cluster with our HR info ... I know in Windows, you just sort of right-click on the node in Failover Manager, and choose "Evict Node". But in SQL to gracefully remove the SQL parts from the cluster, before removing the actual Windows node ... I seem to recall that it's something as simple as remove SQL from cluster by using the SQL setup from the DVD; and choosing remove node #2, https://www.mssqltips.com/sqlservertip/2172/uninstalling-a-sql-server-clustered-instance/ and then evicting node #2 from the Windows cluster. Then I can re-format node #2 as Win2012 R2 and move on from there. 1. Is the procedure in that link valid, for SQL 2008 R2? 2. Can you do this live? By which I mean: I don't need to schedule maintenance downtime, right? I could like do this during regular working hours? No impact to the production cluster, removing a node like that? No restarts necessary? (I've got personal things scheduled for the next like 4 evenings, I'd like to do this during my day, if I could) Advice? Gotchas? Stop by and help out for free on your lunch hour? LOL (that last part was a joke ... but if you're nearby, I could use the company ...)
