iSpot & EML running on backup server

January 8th, 2010 by Richard G

Due to a warning light showing on our main database and file server we’ve moved the hosting of both iSpot and EML on to our backup server. This gave us a chance to test this process and well as allowing time to look at the database/file server.

After spending a significant amount of time documenting the process to transfer the sites to the backup system as well as the process of setting up the actual backup architecture, the move to the backup server was reasonably straight forward, the main issues were as follows:

  1. Swapping the domain names to the backup server can take up to 24 hours to propagate through all domain name servers. This means, potentially, that people can not access the site for that amount of time.
  2. The process of rebuilding all of the images used in the EML caused the backup web server to become unresponsive and so both sites became inaccessible until the server was restarted.

The latter issue should now have been solved with the upgrading of the backup web server and the sites now seem stable. The former issue relating to the swapping of domains between servers will pose the same problems when we come to move back to the main server. There’s not much that can be done about this this time and in practice people should be able to see the changed location much sooner than 24 hours but we are going to look at revising the process of moving to our backup server in case of future need, to minimise the site’s potential down time.

Now all the backup architecture has been successfully set up and tested work can begin again on the development of site features. Top of the list is allowing users to be seeded with a knowledgeable status to recognise users that aren’t experts but do have knowledge, this has is now under way.

One Response to “iSpot & EML running on backup server”

  1. Jonathan Says:

    I see you have just added ‘likely ID’, which will be interesting to watch. However, at the minute it even says an ID is likely when there are no agreements and the person making the determination has the lowest level of confidence ‘it might be this’. Can you place a threshold on the likely ID function so that it only works if at least 2 determinations have been made. Otherwise, I think it is potentially misleading or redundant.

Leave a Reply