SQL Server Performance Forum – Threads Archive

Losing All Resources during power outage.

We have a active/passive cluster config. We had a power outage in our plant for about three minutes. The servers, external array, and network connections where all still up.
The clustered servers and array are plugged into two ups’s. Supposedly all the network connections where not lost. But all of our resources went offline. What could have gone down to cause everything to go offline if we had power to the servers, array and network connnections that were still up? Could it have anything to do with the DC. If the dc goes off the network what issues will that cause?

---

Hi ya, when you say all your resources went offline do you mean in the cluster admin tool they showed up as being offline? or do you mean that the cluster continued to work, but users were unable to access it? or are you no longer able to say whether the cluster was working if you were physically at the cluster at the time of the outage? the DC going offline would be bad too, as the DC tends to be used for authentication and name resolution… what OSdomain are we talking about? using WINS or solely DNS? Cheers
Twan

---

---

2000 Server on dc, soley DNS. Yes the resources in the cluster goup in cluster administrator went offline. The event logs showed that both servers lost their public connection. Looks like the switch could of possibly went down for a little bit as well. So if the actual network connection goes down but you still have the internal your resources go offline?

---

Hi ya, yes if you loose the network connection then the resources will go off line. So if your SQL servers plug into a switch and that switch goes off, then both cluster nodes will fail Cheers
Twan

---

Now is that because you have lost the dc connection as well? If not, why wouldn’t the cluster be able to keep the resources up with the internal communication between the servers still going? Thats the heartbeat between the two is it not? Thanks for the info
dcaldwell

---

Hi ya, both nodes would lose the ability to listen for the cluster ip, and it is that which brings down both nodes. They can still hear each other via the internal comms, but the cluster ip address must be available too. This is irrespective of access to any other devices Cheers
Twan

---

]]>