Disk failure

  • What happens if a part of the disk is unreadable during mining? Does the plot need to be created from scratch? I'm thinking about setting up ZFS filesystem to detect failures. Will detecting failures make a difference for mining?

  • administrators

    @dmp1ce If there are disk errors while reading part of the plot, it's hard to say what will happen, and much of it will depend on the miner being used.

    As for using ZFS - I have a 60TB miner mining against ZFS volumes.


  • @haitch Is only some of the plot read to create a valid block when mining?

    What type of ZFS configuration do you have? I assume you are not using any redundancy.

  • administrators

    @dmp1ce All of the plot files are read, but only 1 specific scoop (1 / 4096th of the file) is read. Depending on wether the plot is optimized or not will determine if those scoops are in one sequential location within the file, or all over the place.

    ZFS is 16 drives in 4 stripes - no redundancy. Touch wood, but haven't lost a LUN yet.

  • @haitch Why did you decide to strip the data instead of having a plot for each disk? Thank you all the information! I have about 6TB free right now.

  • administrators

    @dmp1ce I was originally looking to create very large, fully optimized plot files to minimize the mining time. However I had problems trying to plot a 16TB plot file, so reverted to smaller ones (~4TB), but I didn't bother re-configuring the ZFS layout. In hindsight I should have, but I'm not going to replot it all.

Log in to reply