658089 - apply RHEL6 fixes for production 2.0 graphserver

Status: RESOLVED FIXED

(mozilla.org Graveyard :: Server Operations, task)

Description

[Robert Helmer [:rhelmer]]

A production server was set up for graphserver 2.0 in bug 627446, but we had some setup issues that I could not reproduce in staging, due to staging being RHEL 5.5 and prod 6.0

Staging has been upgraded thanks to bug 657407, so I have a list of things which must be changed in order to get it running:

* rpm-packaged webob too old

rpm -e python-webob
yum install python-setuptools

* rebuild node.js

This is documented at https://github.com/rhelmer/graphs/blob/master/INSTALL#L74

* pull latest code, and pass servername to cache.sh cron job

su - graphs
cd /var/www/html/graphs
git pull
edit /etc/cron.d/graphs
# add servername to cache.sh line like so
*/1 * * * * graphs /var/www/html/graphs/scripts/cache.sh graphs-new.mozilla.org

Comment 1

[Justin Dow [:jabba]]

Would it be safe to puppetize the prod server and add this stuff to the puppet class (stage is under puppet's control already)

Comment 2

[Robert Helmer [:rhelmer]]

(In reply to comment #1)
> Would it be safe to puppetize the prod server and add this stuff to the
> puppet class (stage is under puppet's control already)

Sure, note that the "rpm -e python-webob" is undoing a previous manual change though; we just want to make sure there's only one version of webob installed, and right now need to get that through easy_install to get a new enough version. 

If you want to have puppet do the easy_install and manager /etc/cron.d/graphs I think that'd be a no-brainer.

The install is a little hairy right now; it requires changes to some files in the checkout, but I can work with you guys to streamline the install.

Comment 3

[Dumitru Gherman [:dumitru]]

Changes done in production (graphs1.dmz.sjc1), with the observation that putting those through puppet is kind of a small project and has to go in a different bug.
Closing this from now.