[ https://ovirt-jira.atlassian.net/browse/OVIRT-2252?page=com.atlassian.jira.p... ] Barak Korren updated OVIRT-2252: -------------------------------- Description: "Since we only have one s390x slave, it is currently attached to both the staging and the production CI systems, and while they use separate user accounts, it turns out this is not enough to isolate them from one another.\r\n\r\nThe are several issues that are caused by this configuration:\r\n# Tests that allocate a fixed network port can fail if they are run by both systems at the same time - this happens in practice when sending Python patchs to the '{{jenkins}}' repo because the {{mirror_client.py}} tests start a web server on port 8675.\r\n# The {{mock_cleanup.sh}} script that is being run by one system can time out trying to umount things from a mock environment that was created and is being used by the other system.\r\n\r\n\r\n\r\n" (was: One of the tests of the CI mirrors python package is starting up an HTTP server on a fixed port number. This can cause an issue if the same test is running in parallel on the same machine. This typically cannot be an issue on the CI infra because it never runs more then one STDCI thread at a time on a given slave. This issue does arise specifically on the s390x slave because there the same slave is used both by the staging and the production CI systems. In that case the two systems might be trying to run the same test at the same time, so this can lead to port allocation exceptions. )

The s390x slave is used in parallel by both the staging and the production CI systems -------------------------------------------------------------------------------------

Key: OVIRT-2252 URL: https://ovirt-jira.atlassian.net/browse/OVIRT-2252 Project: oVirt - virtualization made easy Issue Type: Bug Components: CI Mirrors Reporter: Barak Korren Assignee: infra

Since we only have one s390x slave, it is currently attached to both the staging and the production CI systems, and while they use separate user accounts, it turns out this is not enough to isolate them from one another. The are several issues that are caused by this configuration: # Tests that allocate a fixed network port can fail if they are run by both systems at the same time - this happens in practice when sending Python patchs to the '{{jenkins}}' repo because the {{mirror_client.py}} tests start a web server on port 8675. # The {{mock_cleanup.sh}} script that is being run by one system can time out trying to umount things from a mock environment that was created and is being used by the other system.

-- This message was sent by Atlassian Jira (v1001.0.0-SNAPSHOT#100087)