[
https://issues.apache.org/jira/browse/AMBARI-16954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15772420#comment-15772420
]
Gervais Mickaël commented on AMBARI-16954:
------------------------------------------
Hi all,
I've the same problem with a removed service. (Spark...) And the message is not
clear because it says Spark is not checked since last update. But it's Spark 2
which is concerned.. maybe, instead of doing a String.contains
(https://github.com/apache/ambari/blob/84ecbb404280b541f15e2b23068ecd449bdd930c/ambari-server/src/main/java/org/apache/ambari/server/checks/ServiceCheckValidityCheck.java#L140)
to check service, an exact match should be done!
Mickaël
> SERVICE_CHECK Upgrade pre-check does not throw error when its expected to
> -------------------------------------------------------------------------
>
> Key: AMBARI-16954
> URL: https://issues.apache.org/jira/browse/AMBARI-16954
> Project: Ambari
> Issue Type: Bug
> Components: ambari-server
> Affects Versions: 2.4.0
> Reporter: Dmitry Lysnichenko
> Assignee: Dmitry Lysnichenko
> Fix For: 2.4.0
>
> Attachments: AMBARI-16954.patch
>
>
> *Steps*
> # Deploy HDP-2.4.0.0 cluster with Ambari 2.2.2
> # Upgrade Ambari to 2.4.0.0
> # Register HDP-2.5.0.0 version and install the bits
> # Modify configs for some of the service like HDFS, ZK, YARN
> # Start EU
> *Result*:
> EU pre-check does *not* report below error for the three services whose
> config was modified in step 4
> "The following service configurations have been updated and their Service
> Checks should be run again:"
> Upon further investigation found that the pre-check does not work if a
> service check has never been run for a service at all AND reports success in
> such cases
> In other words, the comparison between last config modification time and last
> service check time succeeds if service check never ran at all and the output
> of below query returns empty:
> {code}
> SELECT start_time FROM host_role_command where role = 'HDFS_SERVICE_CHECK'
> AND status = 'COMPLETED' ORDER BY start_time DESC;
> {code}
> In this case when I manually ran a service check for HDFS and retried EU, the
> pre-check caught the mismatch and reported error
> *Note*: I believe we do run service check as part of cluster install, but
> looks like it does not get updated in the DB tables for all services.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
