[Top] [All Lists]

Re: Deployment Strategy Issue

Subject: Re: Deployment Strategy Issue
From: "Diane Larsen [MSFT]"
Date: Wed, 30 Aug 2006 10:43:44 -0700
Newsgroups: microsoft.public.sqlserver.notificationsvcs
See my attempt at the answers below.

Diane Larsen [MSFT]
SQL Server User Education

This posting is provided "AS IS" with no warranties, and confers no rights.
Use of included script samples are subject to the terms specified at

"RDV" <[email protected]> wrote in message 
news:[email protected]
> One deployment strategy is to have the SQLNS instance service (NS$) and
> event providers running on one server (A) and the database on another
> server (B).  My questions are:
> 1) How should SQLNS behave when servers A and B are rebooted?  My
> experience is that if Server A comes up first and can't connect to the
> database it aborts and never starts.
A: Your observation is correct. The NS$ services do not have a dependency on 
the database engine service, and do not automatically try to restart. If the 
database engine instance is not available when the NS$ services attempt to 
start, the NS$ service will fail to start and write an error to the Windows 
event log.
> 2) How should SQLNS behave when just server B is rebooted?
A: The NS$ services on server A will write errors to the event log each time 
the service tries to connect to the database on server B, but can't connect. 
However, when the database engine comes back, it should reconnect. (I don't 
think there is a max connection error....) Note that some subscriptions 
might not get processed depending on the SubscriptionQuantumLimit settings. 
(Basically, you could have lots of events and subscriptions backing up. This 
setting allows you to ignore older events and subscriptions and just process 
newer ones.)
> 3) Is there a way to create a service dependence on the instance
> service on server A against SQL Server running on server B (e.g create
> a service dependence against a process running on another server)?
A: Not that I'm aware of.
> 4) In general are there any best practices for making the above
> deployment strategy work when servers are rebooted?
A: I have a single-server deployment, but here's what I do: I configure 
multiple recovery attempts for the NS$ services via the Services panel in 
Windows. Right-click the service, select properties, and then select the 
Recovery tab. Set the first failure, second failure, and subsequent failures 
values to Restart the Service, and then set a reasonable interval in the 
Restart Service After box.

<Prev in Thread] Current Thread [Next in Thread>