[
https://issues.apache.org/jira/browse/VCL-5?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14147701#comment-14147701
]
ASF subversion and git services commented on VCL-5:
---------------------------------------------------
Commit 1627532 from [~jfthomps] in branch 'vcl/trunk'
[ https://svn.apache.org/r1627532 ]
VCL-170 - option to power off blades after reservation - new reload module
VCL-5 - multiple web servers
vcl.sql:
-modified managementnode table: removed trailing comma at end of KEY IPaddress
line
-modified constraints for semaphore table: added semicolon at end of last ADD
CONSTRAINT line
> multiple web servers
> --------------------
>
> Key: VCL-5
> URL: https://issues.apache.org/jira/browse/VCL-5
> Project: VCL
> Issue Type: New Feature
> Components: database, web gui (frontend)
> Affects Versions: 2.2
> Reporter: Josh Thompson
> Fix For: 2.4
>
>
> Modify the frontend code to allow the existence of multiple web servers
> running in a load balanced fashion. The holdup for this has been due to a
> semaphore lock around the code that allocates a machine for a reservation.
> Without the semaphore, two people hitting the site at (nearly) the exact same
> time can end up being assigned the same machine.
> An idea I came up with to handle this is to only run the scheduling portion
> of the code via the XML RPC API. A single web server would be elected as the
> master. All web servers would call the scheduling part of the code via the
> API on the master, allowing the semaphore lock to still function. If the
> master server goes down, an election process would determine the new master.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
