First, my apologies for the downtime the last couple days.

Second, why?

Went to access my site and write a post, and nothing, zip. Not even an nginx-based "nothing to load here", no IP ping reply, nothing. So I went to the instance management panel - yes, it was powered on. Local console? Unresponsive. Soft reboot? Nothing. Log files for the instance were showing a series of input/output errors. So I pulled the virtual plug. Annoying - but crap happens. Even more annoying and less likely that it de facto wouldn't even give me a basic POST on start or reboot indications or anything on the first start - but it finally came up - with more i/o errors. As if the bootloader or file system had eaten itself. This time though the virtual machine responded to a reboot request from the management interface, and I got a proper start/shutdown - at least proper enough to SSH in, though local console was still broken. Web servers/etc. weren't working, so I tried another reboot from within the ssh shell and lo and behold this time it finally came up.

So, between resource-related reasons why I shifted to a virtual private server (VPS) at the time I shifted off of wordpress - I'd had several lockups due to insufficient memory allocation for what is not a big or active site - plus some odd quirks of resource allocation that meant I'd lost the ability to easily take snapshots, and then this, where it took several reboots for the VPS to come up in a usable condition and the "local" console access was still not working, I decide to take advantage of my preset backups and migrate. I didn't need any more of this stupid B.S.

Simple, right?

If only. You see, I hadn't thought that the Domain name settings had a time to live set to 4 hours. And worse, that there was no way to change them. This meant that I had to wait four or more hours until the internet knew where I'd redirected web traffic, and could thus configure the new web server with a proper SSL cert.

So I did get the new site configured and the database backup uploaded last night, and got the pictures up earlier today.  I also migrated to using a new set of name servers that allow me to set a shorter TTL - and allow me to shorten it further in preparation for cutover.

Lessons learned:

  • Good backups are a lifesaver
  • while it took some time today to be in a position to upload the backed up images, it literally was just a case of reversing the rsync statement used for the backups and providing a new server/key for teh new destination. A few minutes later, we're all done.
  • when migrating the blog, it looks like the information tying nodebb to individual posts isn't stored in the database, or the exported config/content file. All those connections were lost, which is why I ended up deleting the comments as very few were truly significant in the long run and the hasle of cleaning/merging/etc. duplicate topics was just not worth it at all. I'm still not planning to use Disqus, but this is the second fault I've got with the nodebb-based commenting plugin - and at some point I need to figure out why relevant comments do not show up inline in the posts.
  • Something does get stored somewhere, as the handful of articles I messed with trying to reconnect the comments now think they're connected despite the loss of linked topics.
  • When creating a new ghost instance to import the database to, create a unique new user, or your old user settings will not get imported, as a duplicate.

I'll see what I can do about overall cleanup and presenting comments inline.  Also about tracking down where the relevant nodebb topic is stored.

Also - it turns out I've got to get email notifications working for nodebb, but even commenting aside, I really do not want to run phpbb or similar.