Mssql-docker: Reduce minimum memory requirement

[herecydev]

Is there any technical limitation to reducing the memory requirements? I appreciate more memory is great for a production instance, but running several dev instances is really straining resources for no benefit.

[twright-msft]

Yes, currently there is a reason for this. We are working on reducing the footprint requirements over the next couple of releases. Stay tuned....

[coderb]

any progress on this? i'd like to be able to deploy small servers to virtual machines but 4GB memory is a pretty hefty requirement for a database that might be < 100MB big.

[twright-msft]

We are actively working on it. No ETA for release just yet.

[twschiller]

Also a problem for Bitbucket Pipeline CI users, since each service is only given 1GB of memory: https://confluence.atlassian.com/bitbucket/use-services-and-databases-in-bitbucket-pipelines-874786688.html

However, we have been able to get MSSQL to run on BitBucket Pipelines. It will crash/terminate connections if you try to load too much data, though

[twright-msft]

Just a quick update on this issue - we are down to a 2GB minimum requirement now as of CU2.

[justin2004]

@herecydev ,
I made a docker image that works around this memory limitation (policy):
https://github.com/justin2004/mssql_server_tiny

[juliomfreitas]

I think this is a major issue!!!! Thanks for trying this workaround @justin2004 but it didn't work on a bitbucket pipeline.

is there an estimate to lower the minimal requirements?

[justin2004]

@juliomfreitas
what error message did you get from the pipeline?

i was able to run (idling) sqlservr on a VM with 256MB of RAM and 1.5GB of swapspace

[amcsi]

@justin2004 It didn't work for me either.
I don't think I can specify the swapspace in Bitbucket Pipelines.

[justin2004]

@amcsi can you share the relevant portion of the log (with the error)?

[amcsi]

@justin2004 It happened a while back, so there's not that much I can share.

SQLSTATE[42000]: [Microsoft][ODBC Driver 17 for SQL Server][SQL Server]There is insufficient system memory in resource pool 'default' to run this query.

https://gist.github.com/amcsi/81a838a911746c1cbdba3230d760eef5

[justin2004]

@amcsi
given that log it actually looks like there just isn't enough virtual memory to run your query.

the docker image i wrote just bypasses a restrictive policy it doesn't make queries require less virtual memory.

from your log:
Detected 820 MB of RAM

Regular steps have 4096 MB of memory in total
https://confluence.atlassian.com/bitbucket/use-services-and-databases-in-bitbucket-pipelines-874786688.html

are you getting all the memory bitbucket promised?

[amcsi]

@justin2004 I'm just working with the 1GB that I get for Bitbucket services...

and I was going to say that's all that services get, and I could swear that was true 4 months ago, but now I see that apparently I could give more memory to services:
https://confluence.atlassian.com/bitbucket/use-services-and-databases-in-bitbucket-pipelines-874786688.html#UseservicesanddatabasesinBitbucketPipelines-Servicememorylimits

So I guess I could try again, but this time doing memory: 2048.

[amcsi]

Seems like it worked :)