Exos x16 - High head cycle count due to toggle between EPC idle_a and idle_b power states

[chruk]

No need to wait a week, the new idle_b setting works fine.

The problem was caused by a combination of two bad defaults and bad smart design in firmware.

Seagate drives parking after two minutes, but also TrueNAS taking temperature readings via smart every 5 mins, so after 2 mins heads park, then 3 mins later they unpark for the smart read request, rinse and repeat for infinity.

My WD Red drives dont go out of whatever idle mode they in for smart requests unless its a smart test. So much better behavior which I am not sure why Seagate dont do the same. So any setting above 5 mins for idle_b or disabling it will solve the problem, I suggest probably something in between 15 mins and 2 hours. Although realistically 5mins is as good as disabled in TrueNAS since the readings will prevent it from ever hitting idle_b state.

 

[diskdiddler]

I have now ran through the following commands on my 8 drives.

SeaChest_PowerControl_x64_windows.exe -d PD1 --powerBalanceFeature disable
SeaChest_PowerControl_x64_windows.exe -d PD4 --EPCfeature disable
SeaChest_Lite_x64_windows.exe -d PD1 --showSupportedFormats
SeaChest_Lite_x64_windows.exe -d PD1 --setSectorSize 4096 --confirm this-will-erase-data


It appears to have worked, just testing them now, which is _INCREDIBLY SLOW_ on my older test machines. It looks like at least 1 is faulty, potentially 2. I'll know more in 2 or 3 days. It's hideously painful but 4.5 years ago I thoroughly tested my 8TB disks and 0 have failed me by testing first.

 

[SeaWolfX]

I have 4 x Seagate Exos 7E8 8TB (ST8000NM000A-2KE101) and I just learned about this. I have had my drives for about one and a half year on a server that is has very little accessed and I can se the Load_Cycle_Count is already above 45K!! Now I am terrified that these drives will start dying on me.

I have these on a Linux system and I am not sure about what to do. I figure I need to install the SeaChest/openSeaChest tool and change some parameters. Does anyone know how to install the tool on Linux and which parameters I need to change? I have the drives set up in a ZFS 2 x mirrored vdev configuration. Are the operations I need to do with the SeaChest tool safe while the drives are in normal operation in the ZFS pool?

 

[pjc]

I have these on a Linux system and I am not sure about what to do. I figure I need to install the SeaChest/openSeaChest tool and change some parameters. Does anyone know how to install the tool on Linux and which parameters I need to change? I have the drives set up in a ZFS 2 x mirrored vdev configuration. Are the operations I need to do with the SeaChest tool safe while the drives are in normal operation in the ZFS pool?

The two SeaChest_PowerControl invocations directly above your post should do what you need, and I don't think they'll interfere with operation. Just download the SeaChest zip file and there will be statically linked binaries that you can simply run.

Definitely don't reformat the sector size, as it will erase data!

 

[pjc]

I have another question about the load cycle count: why would the load cycle count slowly increase during active use?

I disabled all the power-saving features as shown in post 22, so I would have expected the load cycle count to only go up when there's a power cycle. Instead, the load cycle count is still increasing during use, though very slowly.

During a round of badblocks (constant writing/reading for 8 days) it went from 49 to 63. And then in the following 28 days of hammering the drive with seeks and reads, the load cycle count went from 63 to 78.

Any idea why the load cycle count is still be slowly increasing despite constant activity?

 

[asap2go]

I have now ran through the following commands on my 8 drives.

SeaChest_PowerControl_x64_windows.exe -d PD1 --powerBalanceFeature disable
SeaChest_PowerControl_x64_windows.exe -d PD4 --EPCfeature disable
SeaChest_Lite_x64_windows.exe -d PD1 --showSupportedFormats
SeaChest_Lite_x64_windows.exe -d PD1 --setSectorSize 4096 --confirm this-will-erase-data


It appears to have worked, just testing them now, which is _INCREDIBLY SLOW_ on my older test machines. It looks like at least 1 is faulty, potentially 2. I'll know more in 2 or 3 days. It's hideously painful but 4.5 years ago I thoroughly tested my 8TB disks and 0 have failed me by testing first.

Thanks for writing that down! Helped me greatly.