Press "Enter" to skip to content

Category: Query Store

Why Force Query Store Plans

Grant Fritchey explains the wherefore behind query store plan forcing:

But, what else does Force Plan do for you? What if you never experience bad parameter sniffing (you do, but I’m not going to argue the point)? Is there something else that Force Plan can do for you? Heck yes! The whole point of creating the Query Store was in order to address Plan Regression. What the heck is plan regression? When Microsoft makes any change to the Query Optimizer, and those changes come all the time, it’s possible that you might see a change in your execution plans. Most of the time, it’s going to be a positive change. That’s why they’re changing the Optimizer after all, to improve it. However, sometimes, you’re benefiting from the old behavior of the Optimizer and that new plan doesn’t work as well as the old plan. This is plan regression. When Microsoft changed the Cardinality Estimation engine in SQL Server 2014, more than a few people experienced the new estimator giving row estimates that resulted in a different execution plan that didn’t perform as well as the old plan. This is plan regression. What to do?

This is a good read.

Comments closed

Query Store And Forcing Plans

Andy Kelly explains that forcing query plans using Query Store doesn’t always result in exactly the same plan being used:

Let’s summarize the situation. We have 2 query plans in the Query Store and the most recent one is also the current plan in the plan cache that is being used for all new executions of this procedure. But that plan is bad for all but .1% of the values we may pass in to the procedure. The previous plan in the Query Store is a much better plan overall and that is the one we want to ensure is used regardless of the value passed in. As such we go ahead and force the plan using the provided tools or TSQL which sets the is_forced_plan to 1 for the 1st plan in sys.query_store_plan. As a simplified explanation this action invokes a recompile and the current plan (which was bad) is replaced with a new plan that is based on the one we forced. That new plan now becomes the current one in the cache and is now the one in the Query Store that all new statistics are tied to as well.

Most people would think that if they forced a particular plan that was in the Query Store and it was marked as forced we would in fact be using that very same plan identified by the plan_id and query_plan_hash which is tied to the plan we forced. Keep in mind that if there were problems with the recompile such as it was missing an index that was there when the original plan was created we would get an error which would be listed in the force failure columns and a different plan would obviously need to be used. Errors aside most of the time when we force a plan it gets recompiled and we end up with the same plan as that which we forced. If that plan is the same as the original one we forced it will have the same query_plan_hash and thus the same plan_id.  All future executions will now use that plan and all statistics will be tied to it as well.  This is exactly what we would expect once we forced a plan in the Query Store.

If you’re looking at using Query Store, definitely read this post.

Comments closed

Query Store And Better Plans

Grant Fritchey asks, what happens when you force a Query Store plan and later a better plan becomes available?

Let’s answer a different question first. What happens when the plan gets invalidated, when the index being used gets dropped or some other structural change occurs so that the plan is no longer valid? I answered that question in this blog post. The plan being forced, after the object is dropped, becomes invalid, so that plan can no longer be used. The Query Store still attempts to apply the plan during any recompile or compile event of the query in question, but it fails and a proper plan is used. All this means, I think, the Query Store is going to ignore the new index, since a new index doesn’t invalidate an existing plan. A new index just makes new plans possible. However, when I was asked this question, this wasn’t something I had tested, so I gave a speculative, best guess, answer with plenty of caveats and the promise to provide a tested answer ASAP. Here we go.

Read on for the full answer.

Comments closed

Query Store And Dropped Objects

Grant Fritchey gives us a potential nightmare scenario with Query Store:

Along comes our aggressive junior DBA who decides that there are “too many” indexes on the server. No, I don’t know what that means either, but they evidently read it on the internet or something so they drop the index we created before:

1
DROP INDEX TransactionOccurredWhenNCI ON Warehouse.StockItemTransactions;

What now happens to our lovely execution plan and the plan forcing? We’ll take a look at two events in Extended Events, sql_statement_recompile and query_store_plan_forcing_failed. Nothing happens immediately on dropping the index. The plans associated with that object, if any, are marked as invalid in the cache. The next time we call the query it’s going to recompile and we can see the event:

The result is a bit happier than I would have expected; I was looking forward to a “and the world came crashing down” conclusion.

2 Comments

Query Store Bug

Brent Ozar reports on a bug in Query Store in Standard and Express editions:

If you’re using SQL Server 2016’s awesome new feature, Query Store, there’s a new bug with automatic cleanup. Books Online explains:

Automatic data cleanup fails on editions other than Enterprise and Developer. Consequently, space used by the Query Store will grow over time until configured limit is reached, if data is not purged manually. If not mitigated, this issue will also fill up disk space allocated for the error logs, as every attempt to execute cleanup will produce a dump file.

Here’s hoping that bug gets fixed quickly.

Comments closed

Query Store On Read-Only Databases

Kendra Little wants to know if Query Store works on read-only databases:

At least for now, Query Store can only record query performance in databases that are read-write. Once you go read-only you can review the performance of past queries, but you can’t track the performance of anyone who queried the database after the point it went read-only.

At least for now. Query Store is such an awesome feature that perhaps this will change in the future. (I don’t have any inside info, only optimism.)

That’s a little bit of a letdown, but makes perfect sense.

Comments closed

Using Query Store

Justin Goodwin looks at Query Store in SQL Server 2016:

You can see that our problem query is incredibly easy to find in the top left window based on total duration. Also notice that in the top right Plan summary window, there is currently only one available plan for the query (plan_id 49).

We need to figure out how we can get our ‘good plan’ using Trace Flag 9481 as an available plan that we can force using the Query Store.

The cardinality estimator change in SQL Server 2014 wasn’t perfect, but when you can fix individual plans like this, it makes SQL Server much more powerful.

1 Comment

Query Store Or Plan Guide?

Grant Fritchey answers a big question:

While presenting at SQLDay in Wroclaw, Poland, on the Query Store, I was asked a pretty simple question, which takes precedence, the Query Store or a Plan Guide?

One of my favorite answers to questions is “I don’t know” because it gives me the opportunity to learn. Let’s figure this one out together.

I’ll post the code to recreate this experiment within AdventureWorks at the end of the article. I’m doing this because the code for forcing execution plans using Plan Guides can be pretty doggone long (you may need to generate your own XML from a plan on your own system, fair warning).

The answer is not quite as clear-cut as I would have expected, and I’ll be interested to see what others find.

Comments closed

Forcing Query Store Plans

Grant Fritchey wonders, if you force a plan using Query Store but the plan ages out of cache, do you still use the forced plan?

To start with, a small stored procedure that I use all the time for bad parameter sniffing demos that reliably gets different plans with different values due to statistics skew:

1
2
3
4
5
6
7
8
9
10
11
12
CREATE PROC dbo.spAddressByCity @City NVARCHAR(30)
AS
SELECT  a.AddressID,
        a.AddressLine1,
        a.AddressLine2,
        a.City,
        sp.Name AS StateProvinceName,
        a.PostalCode
FROM    Person.Address AS a
JOIN    Person.StateProvince AS sp
        ON a.StateProvinceID = sp.StateProvinceID
WHERE   a.City = @City;

If this procedure is called for a value of ‘London’ it gets a plan with a Merge Join. For most other value it gets a plan with a Loops Join. Here’s an example of the ‘London’ plan:

It’s a good question with a good answer.

Comments closed

Default Query Store Settings

Erin Stellato talks about default query store settings:

For each database where I enable Query Store, I’d consider the workload and then look at the settings.  I tend to think that the default value of 100MB for MAX_STORAGE_SIZE_MB is really low, and I would be inclined to bump up STALE_QUERY_THRESHOLD_DAYS from 30 to something a bit higher.  I’d also probably drop  DATA_FLUSH_INTERVAL_SECONDS to something lower than 900 seconds (15 minutes) if my storage can support it.  This setting determines how often Query Store data is flushed to disk.  If it’s every 15 minutes, then I could potentially lose 15 minutes of Query Store data if my server happened to crash before it could be written to disk.  I’d also think about changing INTERVAL_LENGTH_MINUTES to a value smaller than 60 if I wanted to aggregate my query data over a smaller amount of time.  Sometimes interesting events happen within a 60 minute time frame, and they get can lost when data is aggregated across that window.  However, aggregating more frequently means I’m adding processing overhead to the system – there’s a trade-off there to figure out.

In our environment at least, 100 MB of query store data would last, oh, a couple hours?  Definitely tweak your settings and keep an eye on them early on.

Comments closed