Go Premium for a chance to win a PS4. Enter to Win

x
  • Status: Solved
  • Priority: Medium
  • Security: Public
  • Views: 605
  • Last Modified:

Cannot use Windows Auth (Kerberos) to connect from Failover Cluster nodes

We recently installed updates on two domain controllers at what I will call SITE-1, of a 4 DC Sites/Services configuration and after it was complete we've started to see some strange behavior (it did not present immediately).  Currently we receive the attached error message any time we try to connect to a Failover Cluster SQL instance from a Failover Cluster node (error_sqlconnect) at both SITE-1 and SITE-2 via SSMS.  We have one node member at SITE-2 that currently holds our 'Production' SQL instance which is working, so luckily we are not hard down because of this, but all replication and mirroring has stopped because of, we think, whatever is causing this error.  We have no tolerance if this single node member goes down, then we would be down and my hair would be in hot lava (already on fire).

We use sites and services in AD to control which DCs can be contacted by which servers and those rules are applying properly as far as logging shows for SITE-1 and SITE-2.  I should mention that the Failover Clusters at both SITES are being affected by this.  Interestingly, when we try and start up mirroring and replication services the service accounts are immediately locked out and SSPI failure messages are logged in our SQL logs.  I also want to point out that ALL other functions of

We have an ongoing case open with Microsoft (Sev A) but so far they have been unresponsive and unable to resolve our case.  It was opened 50 hours ago and we have had 4 hours of engineer time and the rest we have simply been waiting.  Please ask any questions you'd like and I'll try to answer what I can (within reason).

Things we have done:

- Ran dcdiag and repadmin /showrepl with no errors.
- Verified DNS at both SITES and made sure that ALL node members and instances resolve properly.
- SQL validation of connections and checked logs
- Validated SQL connections from other servers that were not cluster members
- Validated that SQL can authenticate fine to DCs
- Confirmed that any authentication from cluster nodes cause accounts to lockout and return the SSPI error in SQL and produces the error attached.
- SSPI context tool used to verify that when accounts don't lock out we could get a successful connection
- Restarted cluster nodes (did not help)
- Used Sites and Services to replicate now between all DCs (both SITE-1 and SITE-2)
- Created new service accounts (these are locking out as well)
-

Things we have not done:

- Restarted SITE-2 DCs
- Restarted active cluster node
- Destroyed clusters
- Installed fresh

As I try other things and find additional information I will add it to our ASK.  Thanks in advance for any assistance you smart masses provide.
error-sqlconnect.png
0
Propay
Asked:
Propay
  • 2
  • 2
1 Solution
 
Ryan McCauleyCommented:
Did the updates you ran (mentioned in the beginning on the article) require a reboot that you've not performed yet? I'm suspicious of those updates, though you did also mention that the behaviour didn't manifest itself immediately after applying them, but some time afterwards - how much time? Since you've got multiple DCs (and presumably multiple GC servers), could you restart them one at a time without any login issues to see if that's related? I'd start there since that's where you applied the updates.

Also, not sure it's the updates, but can you try rolling them back? Just another thought.

And just to confirm the problem, you're getting this error when you connect to any cluster instance from any cluster member (except your one-node cluster)? If you have multiple clusters, do you get this error when connecting to instances on other clusters, or only when  you connect to instances local to the cluster to whose node you're connected? I'm trying to look for some commonality in the issue you're seeing, since it's odd that you're not seeing it on the single-node cluster you have (it sounds like).
0
 
PropayAuthor Commented:
Thanks for the reply.  This actually was caused by a change made over a year ago (no joke) to LAN Manager settings where we allowed the cluster to send NTLMv1 responses but required NTLMv2 everywhere else.  After restarting the SITE-1 DCs, it reared its ugly head as an issue (never seen before).  Once we set the requirement of NTLMv2 on the cluster nodes, like magic it all started working again.  No more locked accounts.  Because you replied, you get the points.
0
 
Marten RuneCommented:
Two questions and one idea.
Q1: what does the errorlog of the instance say, specifically regarding SPN registration?
Q2: do you use ipv6, or only ipv4?
Idea: what happens if you log on using a sql account, i e SA or eqivalent?

Regards Marten
0
 
Marten RuneCommented:
Sorry used my smartphone dis not see replies

Marten
0
 
PropayAuthor Commented:
My comment was a reposting of the solution I found.
0

Featured Post

Industry Leaders: We Want Your Opinion!

We value your feedback.

Take our survey and automatically be enter to win anyone of the following:
Yeti Cooler, Amazon eGift Card, and Movie eGift Card!

  • 2
  • 2
Tackle projects and never again get stuck behind a technical roadblock.
Join Now