Testing the ST4000DM004 on a different machine not using a SATA card and the disk now seems to be working. It suggests the logic on the disk is problematic as other drive managed shingled disks are not at issue. Drive managed shingled disks have had some issues with firmware when they were first introduced and NAS users discovered they would not work properly.
The disk has a 256MB cache which is enough to buffer about one or two tracks. It’s not clear how many tracks make up a given shingled block so its impossible to analyse the bottlenecks in performance.
The disk was formatted with Windows Disk Manager without error. The disk is now being tested with SMART long test to be sure the unit is operational. The WB Lifeguard tool suggests the 4TB disk will take about 6 hours to run the extended SMART test. After testing completed the disk was then fully erased which took another 6 hours. Testing the disks this extensively is necessary to be sure it is safe to use. The full erase allows a disk to activate spare sectors.
The SATA card in the main gaming box has 8 ports which bas based on a 4-port SATA controller and a SATA port expander to increase the ports. The card would not be needed if AMD and Intel would put 10-12 SATA ports along the edge of the motherboard. The incremental cost is minimal. The studio also has 4-port SATA cards which are common and very low cost.
In a recent Backblaze report on media survival they reported that SSD products were far lower in failure rates compared to hard disks. The oldest disks that backblaze have are 8 years old so they are over 70,000 power on hours which is slightly more than disks in the machines here. The Backblaze AFR is 1.49% which is far lower than the studio average. Then again Backblaze did not have the RoHS disk solder problems the 200GB-500GB era suffered. Then there was the Teapo capacitors which added to the woes.
In the transition to RoHS disk failures spiked before solders were refined enough to eliminate problems. Several disks in the studio in the 200GB to 750GB range tended to not last very long.
With the disk back in service the machine can use the disk as a scratch pad for compressing some old archives with 7-zip which are presently only stored uncompressed. Drive managed disks are bets for storing compressed archives. 2TB USB sticks are a new option which can handle cold storage to some extent. 7-zip can be slow like all compression tools so having the efficient machine on the LAN makes it convenient.
Lager scale NAS boxes with more than 4 disks tend to striping data across several disks at once. For some reason the drive managed disks do not seem to work well in the scenario. It seems that drive managed shingled disks are best as secondary storage for media libraries and archives etc.
The disk in now in the SPEC-01 machine and the disk now is being used for archival files such as operating system installers and compressed archives. For the most part the machine is a NAS unit for want of anything.
Additional disks are being considered given the low cost of 8TB hard disks. Checking in with NewEgg shows low cost disks have largely run out. Could be that Seagate is developing a new lineup of disks?
Now that the disk is recovered it has been returned to service after testing showed the disk is operational Adding another 8TB disk expanded the pool and also reduced the average operating hours. There are 9 disk bays in the 750D so it is a fairly large pool by PC gaming standards. The larger population reduces turnover in disks that remain functional. While the table shows 62 disks, this does not show older MFM disks which do not have any way to monitor operating hours. MFM disks were widely used from 1979 to 1988 when the IDE disks became more affordable. IDE disks were replaced by SATA disks which have been in use since. SATA disks dominate the collection primarily as capacity grew they overwhelmed the performance of older disks. 4TB disks for less than $80 is now how affordable storage has become.