Author |
Topic |
kirtigirdhar
Starting Member
9 Posts |
Posted - 2007-08-27 : 04:40:11
|
If the primary master server become permanently unavailable, following a hardware failure for example, what could steps to promote the secondary master to the role of primary master. The Secondary server is a updateable subscriber and there are other servers also in the cluster with read-only subscription. As per my tests if primary goes down, we can create new publication on secondary and create its subscriptions.everything should work fine after that.if primary comes up in any case that can create problem in new cluster so we should take care in this regardplease let me know if something is missing here. |
|
kirtigirdhar
Starting Member
9 Posts |
Posted - 2007-08-27 : 05:46:12
|
just to inform. Replication is Transactional Replication |
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2007-08-27 : 13:44:27
|
There's nothing to promote with clustering. If one server goes down, the instance moves to the other node. SQL Server is a virtual server in a cluster. Replication isn't impacted when a node goes down. Replication moves with the instance to the other node. You don't even need to rebuild replication in this case.Tara KizerMicrosoft MVP for Windows Server System - SQL Serverhttp://weblogs.sqlteam.com/tarad/ |
|
|
rmiao
Master Smack Fu Yak Hacker
7266 Posts |
Posted - 2007-08-27 : 21:43:56
|
If publisher goes down, you need to point apps to subscriber. Sql doesn't promote subscriber like db mirrowing does. |
|
|
kirtigirdhar
Starting Member
9 Posts |
Posted - 2007-08-29 : 02:20:07
|
for information,in our setup we have one primary- one updateable subscriber(secondary ) and other servers are read-only subscibers( one directional). Updates happens on primary only and on secondary if primary is down. reads happens on all the servers. That is the reason why we need to upgade secondary if primary goes down. |
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2007-08-29 : 02:28:01
|
But if you are using a cluster, then there shouldn't be a reason why the primary instance would go down. The instance just moves to another node.Tara KizerMicrosoft MVP for Windows Server System - SQL Serverhttp://weblogs.sqlteam.com/tarad/ |
|
|
kirtigirdhar
Starting Member
9 Posts |
Posted - 2007-08-29 : 02:34:36
|
please excuse me for earlier information , the cluster is NT cluster and we are using sql server replication to replicate data across machine. |
|
|
kirtigirdhar
Starting Member
9 Posts |
Posted - 2007-08-29 : 02:49:43
|
quote: Originally posted by kirtigirdhar please excuse me for any confusion, the cluster is NT cluster not sql server cluster and we are using sql server replication to replicate data across machine.
|
|
|
tkizer
Almighty SQL Goddess
38200 Posts |
Posted - 2007-08-29 : 12:37:05
|
So this isn't a Windows cluster?Tara KizerMicrosoft MVP for Windows Server System - SQL Serverhttp://weblogs.sqlteam.com/tarad/ |
|
|
rmiao
Master Smack Fu Yak Hacker
7266 Posts |
Posted - 2007-08-30 : 01:04:31
|
I believe NT is windows. Why don't set clustered sql? |
|
|
kirtigirdhar
Starting Member
9 Posts |
Posted - 2007-08-30 : 02:00:22
|
yup, we can think of sql cluster. but current system is in production and require solution for that.may be some start-up kind of . not sure what could be done. |
|
|
rmiao
Master Smack Fu Yak Hacker
7266 Posts |
Posted - 2007-08-30 : 02:18:34
|
If you don't have clustered sql, why 'primary comes up in any case that can create problem in new cluster'? |
|
|
kirtigirdhar
Starting Member
9 Posts |
Posted - 2007-08-30 : 02:37:16
|
the scenario is we have one primary(publisher), one secondary(updateable subscriber) and two read only subscribers and Replication is Transactional Replication.let us assume we remove primary from the LAN. we upgrade secondary.we create publication on secondary and we also upgrade one of read only subscriber to updateable subscriber and another read-only subscriber subscribe from new primary.now everything is working fine. After some days old primary comes back in LAN and try to publish or try to sync up with its old subscriber. that can create problem.I hope situation is clear now. |
|
|
rmiao
Master Smack Fu Yak Hacker
7266 Posts |
Posted - 2007-08-30 : 02:49:50
|
So nothing to do with cluster. Just remove publication from it before putting it in LAN. |
|
|
kirtigirdhar
Starting Member
9 Posts |
Posted - 2007-08-30 : 10:25:44
|
i also think this is the only thing we can do in this case. thanks for helping. if anybody else has any suggestion that would be great help |
|
|
jen
Master Smack Fu Yak Hacker
4110 Posts |
Posted - 2007-09-03 : 17:02:59
|
disable the replication jobs in the old publisher? in any case you need to reinitialize everything to get it working again--------------------keeping it simple... |
|
|
|