uPortal IRC Logs-2013-08-01

Owned by Apereo Infrastructure
Last updated: Aug 01, 2013
2 min read

[14:52:36 CDT(-0500)] <tlev> @drewwills1 - so we deployed out 4.0.12 in production and it caused an outage due to a new purge job that deletes unused portlet cookies from the database. Should we document that as a "warning" on the release page or something?

[14:56:01 CDT(-0500)] <drewwills1> tlev it certainly seems like something we should communicate

[14:56:16 CDT(-0500)] <drewwills1> is there anything on the list yet about it?

[14:56:23 CDT(-0500)] <tlev> i do not believe so

[14:57:13 CDT(-0500)] <drewwills1> do you have time to post enough description to allow me to build a decent mental picture of the issue?

[14:57:14 CDT(-0500)] <tlev> I was trying to hunt down the Jira issue that caused it, but i wanted to get others ideas before sending a mass email or updating the release notes

[14:58:10 CDT(-0500)] <drewwills1> I happy to get up to speed on this... make some suggestions

[16:13:40 CDT(-0500)] <tlev> that took way to long to find, but i found it

[16:13:41 CDT(-0500)] <tlev> https://issues.jasig.org/browse/UP-3713

[16:24:04 CDT(-0500)] <tlev> Looking at the commit, it looks like it was something Eric worked on and i pulled into jasig. Most of the work was done on PortletCookieServiceImpl.java

[16:42:28 CDT(-0500)] <drewwills> @tlev I had written myself a note to check in with you around this time regarding your thoughts on a release anyway

[16:45:10 CDT(-0500)] <tlev> When we started up our staging server it locked pretty much that whole table. which was about 20 million records. The fix was to do the delete before the release.

[16:48:50 CDT(-0500)] <holdorph> sounds like a poorly written purge job, if it's locking the whole table

[16:49:57 CDT(-0500)] <tlev> it just locks the rows it needs to delete, but caused an issue

[16:50:05 CDT(-0500)] <tlev> since there was so many

[16:50:29 CDT(-0500)] <tlev> it will be a good solution going forward, but making the transition should have been handled better

[16:50:55 CDT(-0500)] <holdorph> maybe there needs to be a migration sql script provided to people as part of the release notes.

[16:50:59 CDT(-0500)] <holdorph> not just a warning

[16:51:07 CDT(-0500)] <tlev> that would be good, i agree