On a new install of truenas scale with a supermicro x11ssl-cf, I noticed once on reboot that the drive letters changed. Is that normal? I wanted to setup periodic smart tests on the non-ssd drives, but if the drives letters change again, I fear they might run on the wrong drives in the future. I realize that drives can be identified uniquely regardless of drive letters, but just want to be sure it won’t affect any settings if it happens again.
8 hdds are connected via the 2 onboard SAS connections and 2 ssds are connected onboard sata connections.
Yes, drive letters can change. As I understand it, it is first come, first serve. The fasted drive ready to operate get the first position “sda”, the second ready “sdb”. You get the picture. You could be talking about 1 millisecond of difference. That is as I understand it.
I don’t think this will help you solve your problem. Tell me “exactly” what you are trying to do. Be specific. Are you trying to have drive serial number 123 run a SMART short self-test at a specific time, then drive serial number 321 run the same test at a completely different time?
This would be a simple script to write up but you must be specific with what you are trying to accomplish. The way to keep specific drives in order is to use the serial number, not the Drive Name (ID).
If I was on another system and modifying /etc/fstab I would use the unique id of the drive to setup the mount points and not the drive letters.
I was just wanting to confirm that when the drive letters do in fact change, the smart tests I selected based on drive letter will map to the correct drive.
After I posted last night I thought to myself if TrueNAS will keep testing the drives properly, meaning that if I tested sda on Monday and sdb on Tuesday, would it ensure that the physical drive would be tested on the same day regardless of the assigned drive name. My suspicion is no, I believe TrueNAS will test based on the assigned drive name (sda) regardless of if it is the same physical drive or not.
Now this will only happen if you are powering down the server and powering it up. If you leave it on then this never becomes an issue as the drive names will remain the same. And I will tell you that the drive name switching does not happen often, so I have been observing, it should be a minimal impact.
The schedule you listed might be just an example and I realize you and others may have wildly different setups, but for argument sake, your example is immune to this problem (um, free feature) as you are testing all the drives on the same day/time. Now a little feedback on this schedule… If these are all HDDs, I would recommend breaking the testing into two to three groups. The Short tests, you can keep the same days however have three entries to start the Short testing 5 minutes apart and all the testing will be completed in about 17 minutes. The Long tests are a bit different as those tests can take possibly 24 hours or longer if you have 20TB drives, or a few hours if you have 2 TB drives. I would break those apart in similar fashion however chose the same time but different dates, and I would separate these by 2 days only because I have no idea how long your drives take to complete a Long test.
I know you probably did not expect this kind of feedback however I wanted to give you everything I have on the topic. If you have any questions, ask.
That is a bummer, but as you mentioned if running the same types of tests against all drives, it wouldn’t have an impact on ensuring the tests get run.
It is just while I’ve been testing out this new setup I noticed it happening; eventually it’ll just stay on.
In the schedules I did not include the two SSDs that are my mirrored boot drives, but given your suspicion about drive letters and schedules I’ll just setup tests against everything to keep it simple.
As of my post, those were to be my schedules.
These 18TB drives unfortunately take ~28 hours to run the long tests. I was considering splitting them initially and with your suggestion of doing so, I will go back to that. I think the grouping (assuming drive letters ) will be:
Yikes! 28 hours. Yes, you cannot test the drive twice in a 24 hour period.
I always recommend a daily Short test and a weekly Long test if you can do that.
As for the drive arrangement, if you are going to put in the work to do this how I would do this…
Never perform a Long test on two adjacent drives if there is any possible overheating going on due to creating a sort of fire box (drives too close to each other and poor airflow) and to mitigate a hot spot between drives. A lot of high drive temps are due to this poor planning.
Oh wait, that is it really. I have no more advice.
Hope this feedback helps out.
EDIT: The damn “i” key is not working so good. Must be time for a new keyboard.
The 8 drives are stacked in 2x4HDD sections. These new schedules run drives together that are not adjacant. I had in my head that the drives that were vertically higher would be hotter: nope. The ones sandwiched in-between other drives are the warmest. TIL.
) will be:
