[
https://issues.apache.org/jira/browse/VCL-847?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14389379#comment-14389379
]
ASF subversion and git services commented on VCL-847:
-----------------------------------------------------
Commit 1670477 from [~arkurth] in branch 'vcl/trunk'
[ https://svn.apache.org/r1670477 ]
VCL-845
Removed reference to blockRequest.admingroupid from
utils.pm::get_management_node_blockrequests and from DataStructure.pm.
VCL-847
Fixed problem in utils.pm::reservation_being_processed. The
$parent_reservation_id variable was getting initialized every time. Later on it
is set to the actual ID if not defined. As a result, it would never be set to
the correct reservation ID.
Added State.pm::wait_for_reservation_loadstate. Updated reserved.pm so that
child reservations wait for the parent to generate an "acknowledgetimeout"
computerloadlog entry. The timestamp of this entry is used to determine when to
stop checking for acknowledgment.
Updated reclaim.pm to call state_exit instead of switch_state.
Commented out section in State.pm::wait_for_child_reservations_to_exit where it
exited immediately if the request was deleted. This causes problems with
cluster requests. Added a comment explaining the problem for future reference.
Updated how State.pm::state_exit handles setting request.state if the current
state is deleted. It was skipping the request.state update instead of changing
it to completed.
Other
Added code to new.pm::wait_for_child_reservations to use a short interval
between checks at the start, and then increase it. This speeds things up when
all child reservations are ready early on.
> Computer left in reserved state if user deletes cluster request
> ---------------------------------------------------------------
>
> Key: VCL-847
> URL: https://issues.apache.org/jira/browse/VCL-847
> Project: VCL
> Issue Type: Bug
> Components: vcld (backend)
> Affects Versions: 2.4
> Reporter: Andy Kurth
> Assignee: Andy Kurth
> Fix For: 2.4.1
>
>
> The computer of a child cluster reservation may be left in the _reserved_
> state if a user deletes a cluster request during the stage where vcld is
> waiting for the user to acknowledge by clicking _Connect_.
> The loop of the child reservation is calling {{exit;}} if it detects the
> request has been deleted. It should call {{state_exit;}} and pass it the
> argument to set the computer state back to _available_.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
