Tune topology synchronization

StackState Self-hosted v5.0.x
This page describes StackState version 5.0.
Go to the documentation for the latest StackState release.
Overview
This guide lays out steps to tune topology synchronization for best performance on a Kubernetes deployment of StackState.
To tune the topology synchronization performance, follow these steps:
1.​Observe topology synchronization performance​
2.​Investigate StackState platform as the bottleneck​
3.​Change the topology synchronization CPU budget​
This process can be repeated until the desired performance is achieved.
Observe topology synchronization performance
To understand how a topology synchronization is performing, use the stac CLI to take a look at the synchronization's status page. The status page shows the latency of a topology synchronization. This is the amount of time that it took from data being collected at the source, to the moment that the data is stored by the topology synchronization framework. Typically, a lower latency is preferred, however, a higher latency might be acceptable when synchronizing topology from bigger data lakes.
CLI: stac
CLI: sts (new)
1.Get the urn of a synchronization using the stac topology list command.
# List streams
stac topology list
​
        Node Id  Identifier                                                                               Status      Created Components    Deleted Components    Created Relations    Deleted Relations    Errors
---------------  ---------------------------------------------------------------------------------------  --------  --------------------  --------------------  -------------------  -------------------  --------
245676427469735                                                                                           Running                      0                     0                    0                    0         0
154190823099122  urn:stackpack:stackstate-agent-v2:shared:sync:agent                                      Running                 761818                763870              1517959              1519490         0
144667609743389  urn:stackpack:stackstate:instance:44a9ce1e-413c-4c4c-819d-2095c1229dda:sync:stackstate   Running                  13599                  5496                    0                    0       329
2.Get the status page for the synchronization using the urn.
> stac topology show urn:stackpack:stackstate:instance:44a9ce1e-413c-4c4c-819d-2095c1229dda:sync:stackstate
​
        Node Id  Identifier                                                                               Status      Created Components    Deleted Components    Created Relations    Deleted Relations    Errors
---------------  ---------------------------------------------------------------------------------------  --------  --------------------  -------------------- -------------------  -------------------  --------
144667609743389  urn:stackpack:stackstate:instance:44a9ce1e-413c-4c4c-819d-2095c1229dda:sync:stackstate   Running                  13599                  5496                    0                    0       329
​
metric               value between now and 500 seconds ago  value between 500 and 1000 seconds ago    value between 1000 and 1500 seconds ago
-----------------  ---------------------------------------  ----------------------------------------  -----------------------------------------
latency (Seconds)                                   35.754                                    38.120                                    31.274 
⚠️ PLEASE NOTE - from StackState v5.0, the old sts CLI is called stac.
In a future release of StackState, the new sts CLI will fully replace the stac CLI. It is advised to install the new sts CLI and upgrade any installed instance of the old sts CLI to stac. For details see:
​Which version of the sts CLI am I running?​
​Install the new sts CLI and upgrade the old sts CLI to stac​
​Comparison between the CLIs​
Command not currently available in the new sts CLI. Use the stac CLI.
Investigate StackState platform as the bottleneck
To understand whether the StackState platform is a bottleneck when processing topology, we want to know whether the pod that does synchronization has exhausted its CPU resources. We do this using the following steps:
1.Get the configured request for the StackState topology synchroinzation pod
> kubectl get pod -l app.kubernetes.io/component=sync -o=jsonpath='{.items[*]..resources.requests.cpu}'
2
2.Get the consumed resources
> kubectl top pod -l app.kubernetes.io/component=sync
NAME                               CPU(cores)   MEMORY(bytes)
stackstate-sync-665f988dc4-sh4fp   1970m         3234Mi  
In the above example, we observe that 1.970 cores are used by the synchronization pod, where 2 are requested. This means that the pod is very close to its CPU budget and is likely throttled. To remedy this, follow the procedure below to change the topology synchronization CPU budget.
Change the topology synchronization CPU budget
To modify the CPU budget for the topology synchronization, add/change the following configuration items in the values.yaml of your Kubernetes StackState deployment and deploy the change.
stackstate:
    components:
        sync:
            resources:
                requests:
                    cpu: 4
                limits:
                    cpu: 4
To guarantee performance, requests should ideally be set equal to limits. Limits can never be lower than requests.
After making this modification, observe the synchronization performance again to be sure that the changes have had the desired effect. If not, do another iteration to further tune the synchronization.
See also
​Debug topology synchronization​
​Customize values.yaml​
Use tags
Debug topology synchronization
Last modified 7mo ago