You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
When running our Docker environment on ARM it is very slow (unusable) for some workflows. Currently our Docker containers are only available and only built for x86_64, which causes significant performance issues when run on an ARM machine. This is likely due to the I/O performance when running sbt commands.
Describe the solution you'd like
Create a new client image based on ARM that includes the following:
JDK 11
Spark 3.3
sbt 1.5.2 (1.5.x should be fine)
Then test that the client image works by building the Spark Connector, running examples, and running the functional tests. Also need to do the same for the Kerberos version of the client and test that.
We should then integrate the changes into our Docker files, giving users an option to deploy the ARM-based client image.
Describe alternatives you've considered
An alternative would be to test and document the process of running the sbt commands (build, examples, functional tests) outside the Docker containers. Although this may be difficult as all of our scripts and config assume running in Docker.
The text was updated successfully, but these errors were encountered:
Creating an ARM-based client image only resulted in different errors as the versions of the JDK, Spark, and Hadoop were not aligning.
Until there is better ARM support for Vertica and the other images we use, we can use the x86_64-based images on ARM machines, even if they are slower. See #506 for that future ticket.
Is your feature request related to a problem? Please describe.
When running our Docker environment on ARM it is very slow (unusable) for some workflows. Currently our Docker containers are only available and only built for x86_64, which causes significant performance issues when run on an ARM machine. This is likely due to the I/O performance when running sbt commands.
Describe the solution you'd like
Create a new client image based on ARM that includes the following:
Then test that the client image works by building the Spark Connector, running examples, and running the functional tests. Also need to do the same for the Kerberos version of the client and test that.
We should then integrate the changes into our Docker files, giving users an option to deploy the ARM-based client image.
Describe alternatives you've considered
An alternative would be to test and document the process of running the sbt commands (build, examples, functional tests) outside the Docker containers. Although this may be difficult as all of our scripts and config assume running in Docker.
The text was updated successfully, but these errors were encountered: