~5 seconds delay in attachment request

[malte]

Hello,

I am currently dealing with the following problem: We recently moved our Enterprise deployment to new servers and upgraded to 11.5. Since then, requesting attachments via the REST endpoint of a feature service (e.g.  layers/0/features/401/attachments/1) consistently takes around 5 seconds, regardless of the attachment size, which is much slower compared to before. This only applies to the services that reference a postgresql sde, not hosted feature layers. The delay also occurs when I bypass the web adaptor and query via the arcgis server machine directly. Uploading attachments also seems to suffer from the same delay. ArcGIS Server logs the request as taking only a few hundredths or tenths of a second, including the database query, as I would expect, but then it takes ~5 more seconds for the server to actually send the response.

The server runs on Windows Server 2025. I already tried disabling defender temporarily, but no success.

Hope someone can help, thanks in advance.

[George_Thompson]

In the new architecture, is the PostgreSQL geodatabase in the same location as the new 11.5 machines?

Is this in a cloud deployment or local?

Not sure I have an answer but curious.

[malte]

The deployment is on premise, using VMWare. The PostgreSQL is on the same Host as the Enterprise deployment, but on a different VM. It's the same configuration as before the migration.

[George_Thompson]

What do the resources on host look like?

After the upgrade, is the deployment using more resources (RAM / CPU)?

With that delay on the server side, it seems that it is waiting for the server machine to respond.

Random question: are the 11.5 machines fully patched?

[malte]

The host resources look fine, plently still available.

We did have some issues after the migration because of low virtual memory on the arcgis server machine, but this has since been resolved and did not solve the issue.

All enterprise components are fully patched.

[George_Thompson]

Thanks for that information. This may be a case to work with technical support on a deeper investigation.

[David_McRitchie]

One thing I have seen add to response times are change tables. If a service has been up for some time, and had a lot of edits in its lifespan it could have grown large, and be less reactive.

Do you get the same response time if you publish a new test layer and try to request an attachment?

Ruling out if this is database specific or service specific will help address the problem.

Hope that gives some further help,

David

[malte]

Hi David,

I now had time to test a bit more, but without success. I created a new VM, installed the latest supported postgresql version, created a new SDE and published a new service, still the same problem.

I've now created a ticket with technical support.