Coins to go offline during week of September 4

[Steve Sokolowski]

On September 4, 5, and 6, Chris will be installing some new disk controller cards in the coin daemon servers. During this time, about 30 coins will go offline at a time while the disks are reformatted and the data is copied over to the disks on the new cards.

The reason for the new cards is that we discovered that the MegaRAID cards that had come with these servers are 20 years old, and therefore they were released before SSDs were conceived. They don't have support for commands that SSDs need to use to perform well, and over time the SSDs have slowed down to the point where some coins go into error waiting for the disks to respond to calls. By replacing these cards and reformatting the disks, the latest commands will be supported and we expect performance to improve by a factor of 10.

No action is necessary by customers; however, customers that are mining static coins, or who are using PPLNS or solo modes, will see their coins go offline for a few hours on one of those days. Two weeks from now, we expect the orphan rates of these coins to fall significantly, and for coins to be much less likely to go into and out of error due to poor disk performance.

[Steve Sokolowski]

This maintenence has been moved up to today, since the cards were delivered in the mail more quickly than we thought and there is no reason to wait. Chris expects to begin at around 3:30pm EDT.

[Steve Sokolowski]

About 50% of the work has now been completed. Chris is working on the remainder of the servers today. All coins are currently online - both on the completed disks and on the disks that have yet to be completed.

[Steve Sokolowski]

Chris has now moved all coins to the TRIMmed disks, so the errors where coins go into and out of error due to poor disk performance are resolved. Chris still needs to move some operating system disks, and then he will leave Philly later today.

[Steve Sokolowski]

All of the operating systems have been restarted, the old cards have been removed, and this maintenance has concluded.

An examination of the number of coin errors over the past three hours shows that there were no coin errors caused by poor disk performance. The only errors were caused by coins having too few connections, which is something that we unfortunately can't control because many of the networks are weak.

Thanks for your patience while we performed these upgrades, and enjoy the more stable coins!