Status as of Wednesday, November 29, 2017
Posted: Wed Nov 29, 2017 8:57 am
Good morning!
- After a long process that began on July 9, today marks the beginning of a new era for this project. We have finally finished our previous commitments, and Constance Hilliard will be joining in 10 minutes to start pushing forward to a new level. While Constance is extremely capable, training her, as with anyone, is likely to take a significant amount of our time, so it's possible that support times will actually decline before they improve tremendously. In two weeks, there will be about 2.25 times the manpower there was compared to early November.
- As an example of the absolutely ridiculous growth of this industry, yesterday's 45-minute share delay was caused by another exchange taking too long to respond to our price queries. That's the second exchange to be removed from trading in three days, after we axed ccex earlier in the week. I'll be working today on a new price updating algorithm that will gather prices from all exchanges in parallel, and then, after 15 minutes, kill any requests outstanding and send the prices to the mining server. This change will allow exchanges that cannot keep up with the rapid growth to be removed from the system automatically and return when they have more capacity. We apologize for the problem, but every issue like this allows us to stabilize the system even more because this particular problem will be fixed and not happen again.
- We have changed our plans for this weekend's downtime. While there will still be downtime, we are going to rollout the parallel mining server instead of Postgres 10, simply because it is closer to completion. We will probably have to continue to restrict connections because the database will then become the limiting factor. The initial plan is to put up two servers and limit them to 6000 connections each, see what happens for a week, and then upgrade to Postgres 10 and reopen registrations. The long-term plan will be to have some sort of automatic load balancing, but we have no idea where to start with that and it will take a lot of research.
