R80.20 - Is the REST API working asynchronous?

[KonstantinS]

Hi all,

We are currently doing a lot of stuff with the REST API (v1.3).

In general we are sending four API calls:

1. login

2. set-% (e.g. set-access-role)

3. publish

4. logout

Sometimes it happens that the change will not get through and/or the logout will not work.

It seems that the API Server is executing them asychronously, because it helps if we are inserting some waits (e.g. 10sec) between the calls. Without the waits it would be executed within milliseconds and leads into inconsistent objects. Of course, this isn't a solution because in our PROD environment we may have a higher load on the system later.

Has anyone the same or similar issues? Is the API Server executing the calls in some queue?

Regards,

Konstantin

[PhoneBoy]

How many commands are you doing before you publish?
One thing we recommend as a best practice is to limit the number of changes made per session (a few hundred) before you publish.

[KonstantinS]

@PhoneBoy 

Only one command / one change.

[PhoneBoy]

What are the specs on your management server(s) (RAM, CPU)?
If it's multi-domain, how many domains?
Perhaps some tuning needs to be done on the API server for better performance and/or more RAM/CPU may be needed.

[KonstantinS]

It is a fresh installed Management Server with one domain.

We have only the standard objects in place.

[PhoneBoy]

Amount of RAM/CPU in management server?

[KonstantinS]

32GB RAM and 16 CPUs

[Timothy_Hall]

Please see the following SK, your API server may not be flat-out crashing but if it is constrained for memory it can cause some problematic effects.  Heavy use of the API will almost certainly require giving the API server more memory:

sk119553: Security Management API server crashes under heavy load

[KonstantinS]

Hi @Timothy_Hall ,

Thanks, I saw this article, but I don't think that adding one object to an existing Access Role is known as "heavy load".

[PhoneBoy]

I highly recommend getting the TAC involved to troubleshoot this.

[KonstantinS]

Yes, we did already and it is still ongoing.

But I have the hope that someone ran into the same problem.