There is insufficient system memory in resource pool 'default' to run this query. on sql

126,622

Solution 1

SQL Server will start with as much memory as needed, then slowly ramp up until it uses all allocated to it in the Server Properties:

enter image description here

It will not release any of this memory until the service is restarted; this is by design.

It is generally recommended to leave 2ish GB for the OS, and you need to be mindful of any other processing running on the same server as SQL. It is usually recommended to have SQL Server on it's own server without anything else running there.

That said, 1.5 GB of RAM for SQL Server isn't that much. If you don't have more available, it may be time to add some or upgrade the server.

See also: Technet, Brent Ozar

Solution 2

I am posting this answer because someone may find it useful.

You can set max server memory even with this query:

sp_configure 'show advanced options', 1;  
GO  
RECONFIGURE;  
GO  
sp_configure 'max server memory', 32768;   -- for 32 GB
GO  
RECONFIGURE;  
GO

Also you can use this query to release memory without restarting the service simply by setting to 2 GB (2048) and changing again back to 32 GB (32768) or to the value you want.

Please, do not set it below 1 GB because it will cause errors executing queries and therefore you will need to restart the service to be able to execute queries, even this query to increase memory again.

Solution 3

In our case it was because of Memory Optimized table types, with huge amount of data. There was multiple calls to different stored procedures at the same time and each using the same table type and loading huge amount of records in it (>100,000). For our application, there was way to reduce the number of records inserted in to memory optimized table type i.e. instead if storing all selected items in a memory optimized table type we conditionally stored only the non-selected records.

CREATE TYPE [TestType] AS TABLE (
    [TestOrder]    VARCHAR (1500)    NOT NULL,
    [TestDepartment] UNIQUEIDENTIFIER NULL,
    [TestCourse] UNIQUEIDENTIFIER NULL,
    [TestStudent] UNIQUEIDENTIFIER NULL,
    INDEX IX_Test NONCLUSTERED (TestOrder))
    WITH (MEMORY_OPTIMIZED = ON);
Share:
126,622
Tolga Evcimen
Author by

Tolga Evcimen

I am a freelance software engineer working with enterprise clients through Toptal. Currently focused on AWS services and integrations working with a great team for Deckers. -- "Everyone knows that debugging is twice as hard as writing a program in the first place. So if you're as clever as you can be when you write it, how will you ever debug it?"

Updated on April 01, 2021

Comments

  • Tolga Evcimen
    Tolga Evcimen about 3 years

    I have a running service that gets 50-100 queries per minute. And these are not high cost queries. This service has been running for around 3-4 months without any errors.

    Suddenly few days ago it started giving "There is insufficient system memory in resource pool 'default' to run this query." error occasionally. When I investigate the problem I saw that sqlservr.exe was using ~1.5 gb ram and %25 of CPU(all of 1/4CPU). And when I restarted the sqlservr.exe the ram starts from ~50mb and slowly increase till it becomes ~1.5gb then leads to crashes in the apps using it.

    After I have made little bit of research I figured that it is caused by the edition of sql server I use. It was express edition limiting the numbers to those. So I have upgraded my sql server from '2008r2 express' to '2012 enterprise'. When I started the service I thought my problems are finally over, since the service uses only ~60mb of memory, but in an hour unfortunately same problem started occurring, but this time the used memory I see on windows task manager is still ~60mb, not excessing any limits.

    I use EntityFramework as ORM in a wcf service. And along with it I have SqlQueryNotification(broker and stuff) system for some caching operations.

    Am I missing some crucial configuration points? Or 6gbs of memory and my 4 CPU is really not enough for this? But it can't be that because same load was like that for 3 months and there wasn't any error back then, and there is not any change of codes either.

    ram allocated

    ram usage