Moving_Redmine_to_OpenShift

Moving Redmine to OpenShift

At this time this is only an investigation into the feasibility of doing this, not a sure thing.

Summary

The current Redmine instance is sometimes excruciatingly slow (page loads exceeding 15 seconds), and periodically we seem to get a 502 proxy error between thin and Apache. It would also be great to get out of the business of managing the VM hosting Redmine. For this reason, we are interested in exploring whether moving our Redmine instance to OpenShift would be a better solution.

Owner

• Currently assigned to Matt Wagner (matty_dubs), but I’m more than happy to share and/or hand this over

Current status

This task (#3106) currently represents only an investigation into whether this is a good alternative and how much work it would entail.

Screencast Demo

This task doesn’t really lend itself to automated tests. But we could show:
* Quick walkthrough of the OpenShift process, interface.
* Demo of a test Redmine instance running on OpenShift.
* Could show speed test before and after; Pingdom’s tool is reasonably accurate

Implementation tasks

Currently, only #3106 is entered. We can enter sub-tasks if we decide to proceed further.

• #3106 - Investigate OpenShift as an alternative

• Perhaps launch a test (empty?) Redmine instance onto OpenShift

  • (jay) Directions here: https://github.com/openshift/redmine-openshift-quickstart

• Evaluate comparative performance, stability

• If warranted, draw up migration plan

Detailed description

• SSL support?
  • We do have a wildcard cert.
  • SSL is supported, but it doesn’t work with CNAMEs - we cannot upload custom certs.
    ***** Open feature request here: https://www.redhat.com/openshift/community/content/custom-ssl-certificates
    ** Thinking we might run it on redmine.aeolusproject.org - but existing links would break
    *** CNAME support works fine - I have https://redminetest-mattydubs.rhcloud.com/ running and aliased to http://redmine.n1zyy.com/
    *** SSL does not work with CNAMEs , though
    ** Investigate performance comparison
    *** Seeing page load times of 23 seconds: http://tools.pingdom.com/fpt/#!/qNYxuXGSb/https://redminetest-mattydubs.rhcloud.com/projects/testproject- acceptable IMHO, but not great
    *** When I went to do a comparison for the existing site, we timed out after 35 seconds, so 23 seconds, while not great, is certainly an improvement…
    *** Site graphics/styles are not cached by default~~- if we can use .htaccess to allow them to be cached, or move them off to S3 or something, we might see a decent speedup
    ***** This was easily fixed by dropping a .htaccess file in public/.
    ***** I’m also using CloudFront for testing - it acts as a caching proxy… This was as easy as creating a CloudFront deployment and setting asset_host in production.rb.
    * Backup plan for database
    *** You can run rhc app snapshot save -a redmine to snapshot the Redmine app, including files and the database
    *** There is a restore command; I have not tested it yet
    *** Snapshots will briefly stop the application; this means something like 60 seconds of downtime
    ** Does it make sense to look at upgrading Redmine in the process?
    * Can we send emails from OpenShift? Redmine sends out notification emails, etc.
    *** Yes, openshift supports sending emails, not sure on the blacklist business
    *** https://www.redhat.com/openshift/community/blogs/outbound-mail-ports-are-now-open-for-business-on-openshift- strongly implies we should not send mail directly
    ***** Maybe use something like Sendgrid, which is free for up to 200 emails a day: http://sendgrid.com/pricing.html ???
    ** It looks like we’re given 1GB of disk space
    *** Hopefully ample for file uploads, etc.
    *** The FAQ implies that we can request more if needed
    ** See https://www.redhat.com/openshift/community/kb/kb-e1008-persistant-storage-for-openshift-express-applications re: persistent storage~~- special care needed
    • I’m just saving things in data/ up a level as the above suggest — data now persists

Future plan

It’s unclear whether or not this is / should be related, but a few other maybe-relevant ideas:
* Move the static aeolusproject.org site to GitHub Pages (so deploys just take a git push, making it easier for a wider range of people to update the static site)
*** With that in place, I think we would be done with the existing VM we were using and could shut that down
** Upgrade Redmine / switch to the newer fork
*** Can we tweak the interface to be slightly more attractive?
** It boggles my mind that this isn’t already the case (or maybe it is?), but it might be worth looking at setting up a cache for rendered pages, since they’re seemingly slow to render (for some reason) It seems like this is already the case; it’s just that the whole app is slow right now
* Pretty well removed from the migration itself, but it would be nice to find a Markdown plugin for Redmine. Who uses Textile?!
*** There is one, though I made a mess of trying to install it and haven’t revisited since it yet
** Is there any reason we should have a repo browser in Redmine if it’s visible on GitHub? It feels like needless complexity, and pollutes search results (both our own search and on Google!) with dozens of irrelevant results.
*** iirc, this was added before we moved the code to github, though the idea was integration. IOW, if you reference an issue in the commit, it gets hooked to the redmine issue. I am not sure if it also would close said issue if we had the right config, but that was the general idea. That said, I dont feel super strongly about keeping it
** If OpenShift works out for us, we might want to look at moving Justin’s Piwik install over to OpenShift as well to keep things centrally-managed (and to free him from having to host it)

References

• OpenShift Express
• ChiliProject (fork of Redmine with more active ongoing development)