Improve startup time for tests using HiveMinioDataLake#14561
Improve startup time for tests using HiveMinioDataLake#14561findepi wants to merge 1 commit intotrinodb:masterfrom
Conversation
80cb374 to
bd008fe
Compare
There was a problem hiding this comment.
NIT: maybe withHdfsAndHiveRuntime(false) or withHdfsAndHiveRuntimeDiasbled()
There was a problem hiding this comment.
What about using a dedicated HMS docker image instead of stripping on the fly the services?
https://hub.docker.com/r/starburstdata/hive-metastore may be a viable alternative.
There was a problem hiding this comment.
There is also the effort of @nineinchnick to get multi arch support on the existing images. See trinodb/docker-images#143
Not all images include multi arch support but hive3.1-hive has, I guess we can use that image easily for the Iceberg and Delta tests.
There was a problem hiding this comment.
What about using a dedicated HMS docker image instead of stripping on the fly the services?
Good idea, if someone gonna maintain it.
We need one per every hive config we use in these tests, see
io.trino.plugin.hive.containers.HiveHadoop#DEFAULT_IMAGE and io.trino.plugin.hive.containers.HiveHadoop#HIVE3_IMAGE
|
4cae5cb to
4ee3970
Compare
`HiveMinioDataLake` uses `HiveHadoop` only to have the metastore service. Tests using MinIO don't want to use the HDFS and don't need to wait for it. The startup time is especially severe on Apple M1 chips. The change brings down container startup time from ~43s to ~13s.
5feade6 to
676d601
Compare
|
(just rebased to remove commits extracted to #14742) |
HiveMinioDataLakeusesHiveHadooponly to have the metastore service. Tests using MinIO don't want to use the HDFS and don't need to wait for it.The startup time is especially severe on Apple M1 chips. The change brings down container startup time from ~43s to ~13s.