Honestly, I have no idea.

Maybe tweaking the linked server / provider settings could make a difference, but I have no idea.

-- In server A:

CREATE VIEW V_hidden

AS

SELECT *

FROM MyTablesThatLiveInServerA

GO

-- In server B:

CREATE VIEW V_hidden

AS

SELECT *

FROM MyTablesThatLiveInServerB

GO

-- In server A:

CREATE VIEW V

AS

SELECT *

FROM OPENQUERY(B,'SELECT * FROM MyDB.dbo.V_hidden')

GO

-- In Server B:

CREATE VIEW V

AS

SELECT...

Ok, so basically when you issue

SELECT * FROM V -- from server A

what you need to do is query B.db.dbo.V

In this case, change the view definition on server...

I'm under the impression that the view doesn't get passed through correctly.

Is this used in combination with other local tables or just a plain SELECT * FROM B.db.dbo.V ?

Does the error message change now when using the view?

Look in the plan XML for parameter list.

Here's an article on the subject: http://sqlperformance.com/2013/08/t-sql-queries/parameter-sniffing-embedding-and-the-recompile-options

Does the following work from server A?

SELECT * FROM OPENQUERY(B,'SELECT * FROM db.dbo.V')

You can get bad plans for a number of reasons.

Stale statistics, to name one.

Parameter sniffing is another common cause.

Let's see if I understand correctly:

1) You have a view named "V" on server A and B. The view definition is the one you posted.

2) You have linked servers from...

I don't see any of the objects mentioned in the error message, nor the linked server script.

Execution plans are cleared from cache when you make mirror the principal.

See for reference: http://www.sqlsoldier.com/wp/sqlserver/doesamirroringfailovercleartheprocedurecache

Duplicate post.

Replies here please: http://www.sqlservercentral.com/Forums/FindPost1653581.aspx

Diff backup includes all pages modified since last full backup. Was the page touched by reorganize? Then it must be included in the diff.

If it gets too big, take a...

I assume that you are experiencing bad parameter sniffing and get inefficient execution plans.

Generally speaking, updating stats (with 100% sample for skewed data distributions) fixes it (for a while). Permanent...

Show us some code and linked server script please.