Server reboot at approximately 10:10 AM Eastern.
We are rebooting the server to temporarily lower the RAM that ElasticSearch uses. This will help verify if the upgrade yesterday (to use jemalloc) will help the memory problems.
What this means: before jemalloc, memory would grow forever, but now it should grow for a while and then level off. But what if it levels off /above/ the ram that’s available? Since resizing the droplet isn’t an option (due to recent problems on DigitalOcean), lowering ES ram is an easy path forward.
The consequence of this is slower search performance, which I apologize for.
The instance has enough money on the Patreon to pay for a larger droplet, with more ram! I just don’t have the time right now to coordinate with DigitalOcean tech support to resize the droplet, probably causing severe latency issues, so they can diagnose the problem. Once I move, I promise it’ll be my top priority here. Probably mid-September.
I really appreciate your patience. Reach out to @ashfurrow with Q’s.