Bob Proulx wrote: > Joshua Hoblitt wrote: > > The checksum on the numdiff 5.8.1 tarball has mysteriously changed. This > > package was frequently download by a deployment script that is run > > frequently at my $day_job. We have disconnected usage of it as of a few > > moments ago. It is likely this change happened occurred in last day or two. > > Thank you for the report. I can confirm the current file information:
The problem was a faulty SSD in a RAID10 set of four. It was returning corrupted data and reporting it as good data. The faulty drive has been removed from the array. All of the systems have been booted back up and are online again. Things came back online just a few minutes ago. I expect the FSF admins will close the ticket you were CC'd on and you will get an update there when they get the chance. It is late on Friday there. Data being written is written to both good and bad drives in the RAID1 mirror. Data read is interlaced between the good and bad drives. That is why the read data being accessed would sometimes have corruption. By removing the corrupting drive it is believed that all of the data on the remaining good drive is okay. Thank you very much for reporting this problem! Your report was the earliest report by a day and allowed us to jump on the problem soonest before it had a chance to become worse. THANKS! Bob
