Significal Project Analysis time degradation on different setup

[Oleksandr_Haimanov1 (Oleksandr Haimanov)]

Hi,
We migrate from one infra to another one and faced with issue that Project Analysis time increased significantly.
Java project 90k LOC, cpu and memory limits, plugins and Quality Profiles are identical.

SRC set up:
Docker
SonarQube 7.9.1
Avarage duration 30s

DST set up:
Kubernetes
SonarQube 7.9.1
Avarage duration 40 min

[Tobias_Trabelsi]

Hi @Oleksandr_Haimanov1 ,

welcome to the community

can you please clarify where you see the increase? is in on the actual scanner run or in the computation of the results in sonarqube?

also: do you use postgresql and move it as well by any chance? if yes, please try to run VACUUM FULL of your database as documented in our upgrade guide. maybe this will already resolve your issue.

[Oleksandr_Haimanov1 (Oleksandr Haimanov)]

@Tobias_Trabelsi

can you please clarify where you see the increase? is in on the actual scanner run or in the computation of the results in sonarqube?

Hi Tobias, in the computation of the results in sonarqube.
Yes we use postgresSQL it is fresh one. I also try to use embedded H2 and recive the same result

[Tobias_Trabelsi]

okay do you see high load on your SQ pod or in your database during computation and is your database in the same k8s cluster as SQ?

the embedded H2 performance is depending on the IO performance of the underlying filesystem, so if this one performed bad as well you might want to take a look at the node storage. maybe you have a bottleneck right there.

[Oleksandr_Haimanov1 (Oleksandr Haimanov)]

@Tobias_Trabelsi

okay do you see high load on your SQ pod or in your database during computation and is your database in the same k8s cluster as SQ?

Yes database hosted in the same k8s cluster. No any cpu or memory high loads during computation.

@Tobias_Trabelsi

on the IO performance of the underlying

Do you have any suggestions how properly check IO performance?

[Tobias_Trabelsi]

tricky question. i am not aware of a tool that would do this directly in k8s (prometheus node exporter has some io metrics on nodes but not on persistent volumes as far as i know). what you can do thou is you build your own image and load it with dd. then deploy it the same way as sonarqube and exec into the container to run some manual tests. there is a nice article from thomas krenn that might help you with that.

[Oleksandr_Haimanov1 (Oleksandr Haimanov)]

@Tobias_Trabelsi

there is a nice article from thomas krenn that might help you with that.

Thanks I will try and after that will back to you

[Oleksandr_Haimanov1 (Oleksandr Haimanov)]

Thanks for help. Looks like issue was with IO performance. We upgraded to premium SSD and now it looks good. Topic can be closed.