[
https://issues.apache.org/jira/browse/CASSANDRA-18051?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17679301#comment-17679301
]
Andres de la Peña commented on CASSANDRA-18051:
-----------------------------------------------
{quote}XLarge spends double the credits we spend for Large per minute
{quote}
One might think that the larger resources would finish in less minutes, but
that doesn't seem the case:
* This old [MIDRES
run|https://app.circleci.com/pipelines/github/ekaterinadimitrova2/cassandra/2130/workflows/9df2d073-c26a-4750-82e4-8b6645f37f9c/jobs/16568/parallel-runs/1?filterBy=ALL]
of large dtests that I have stolen from [~e.dimitrova] seems to take ~93
minutes when adding the times of its four runners. It uses large containers.
* This [HIGHRES
run|https://app.circleci.com/pipelines/github/ekaterinadimitrova2/cassandra/2129/workflows/b451fdf4-e643-42e9-86a4-f849e57c559f/jobs/16685/steps]
takes ~94 minutes to finish with its only runner. It uses xlarge containers.
Both runs contain 42 tests. I assume that the minutes we pay are the sum of
minutes we see on the GUI for every runner.
So it seems that the larger resources double the price while keeping the total
time unchanged.
> Optimize test_network_topology_strategy and
> test_network_topology_strategy_each_quorum for large containers in CircleCI
> -----------------------------------------------------------------------------------------------------------------------
>
> Key: CASSANDRA-18051
> URL: https://issues.apache.org/jira/browse/CASSANDRA-18051
> Project: Cassandra
> Issue Type: Task
> Components: Test/dtest/python
> Reporter: Ekaterina Dimitrova
> Assignee: Ekaterina Dimitrova
> Priority: Normal
>
> As pointed in CASSANDRA-18001, there are two Python Large DTests that fail
> because of not enough resources with large containers. [~adelapena] looked
> into them and suggested that the tests themselves can be optimized for
> running on Large containers instead of using XLarge containers for only two
> tests from the whole suite.
> {quote}As for the two tests starting nine nodes each,
> {{test_network_topology_strategy}} and
> {{{}test_network_topology_strategy_each_quorum{}}}, I think they were added
> by CASSANDRA-10584 to test multi-DC consistency levels. I wonder if they
> actually need to start 3 data centers with 3 nodes each. Would two data
> centers with 3 nodes be enough to test multi-DC consistency levels with
> multiple data centers? 6 nodes should be doable on Circle with the current
> resources.
> {quote}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
