I just discovered the concept of TLER / SCT and set these for my HDD RAIDZ1 pool.
I was wondering what values I should use here. I plucked some numbers out of thin air based on my limited knowledge of HDDs and RAID:
Read timeout 2.0 secs - we have redundant disks so if a single drive has a failed read, I want it to skip quickly and use the redundant drive to work out what that block should have been. But is 2.0 secs the best value or would 0.1 secs or 5.0 secs be better?
Write timeout 10.0 secs - I don’t want the NAS to spend forever writing a block, but I do want it to have a decent attempt. But is 10.0secs the best value?
How do these values impact alternate block allocation by the drive firmware? Do they help or hinder?
Good questions. Most people don’t care too much about these details and just go with the sane defaults of NAS-grade disks, which are “a few seconds” (5? 7?).
Probably drive/vendor specific. I’ve never seen anything written, if you do, please share.
I would expect that NAS drives have sane values for the intended use.
My second backup NAS runs on a mix of NAS drives and shucked white label HDDs. When our former Resident Grinch posted a resource about TLER, I set the white label drives to match the NAS drives and didn’t look back.
The default for the Ironwolf drives was Disabled. Obviously there is going to be a finite timeout, so I suspect this means Default, but I have no idea what that default is.
I now have multi-reports implemented that does this, but that runs once a day and the SmartCtl runs once every 5 mins. So this is a better solution - I will wait to see if I can get any more feedback on the default values or my guesses before doing this as at the moment the settings are system-wide as part of multi-report and they will become device by device and more hassle to change once I have done this.
I just use NAS disks as I believe they have sensible TLER values and also, I’ve come to believe the multi-disk rotational vibration sensors are actually a thing.
