Cyknight
Dabbler
- Joined
- May 2, 2015
- Messages
- 19
So a bit of a weird one here that I can't find information on.
I'm currently just stress-testing my new build as per my signature. No data on there that isn't already a duplicate, there is no concern about data loss.
On the zpool I have that runs 9 HDD in RAIDZ3, I'm having an odd issue with ONE drive.
Running what would be normal write/read operations it runs fine, no reported issues.
Yet sometimes (but not every time), when I force high simultaneous writes/reads to it, it goes to FAULTED with 300-800 write errors.
Don't see any cam errors.
If I try to bring it online, it'll resilvers every time, no issues. Subsequent SMART test comes up fine.
In a short testing I found that it's fine if I only do writes or reads. It only fails if it's doing both at the same time.
So, figured bad drive, no problem - swapped it with an other older drive, and it gives me the same issues.
For giggles, I figured I'd try swapping connections, just to see, and the issues follows the drive, so one would think it's not the connector (even swapped which port on the HBA was used just in case)
Two bad drives? Well they're both been in use previously so sure, certainly possible (they're completely different manufacturer and date mind you)
Just because I'm only testing anyway, I figured, let's mess around and put both of the drives that give me errors in the pool....
Here's where it gets truly odd.
I still get the same error, still only on combined reads and writes, but only on one of them at a time. They pretty much go 50/50. But no matter how hard I try, I can never fault both drives at the same time. Indeed I cannot fault the second drive even when it falls or is put offline completely.
Now, in the end, I'm just going to get a new drive and put it in at this point, but I am curious if anyone's seen this before, and if so, if they know a cause - especially if it's NOT an HDD failure I'd love to know before I bring this build 'live' (probably a week, maybe two, away still as of posting)
Oh, and I'm running latest FREENAS stable, no nightlies here.
I'm currently just stress-testing my new build as per my signature. No data on there that isn't already a duplicate, there is no concern about data loss.
On the zpool I have that runs 9 HDD in RAIDZ3, I'm having an odd issue with ONE drive.
Running what would be normal write/read operations it runs fine, no reported issues.
Yet sometimes (but not every time), when I force high simultaneous writes/reads to it, it goes to FAULTED with 300-800 write errors.
Don't see any cam errors.
If I try to bring it online, it'll resilvers every time, no issues. Subsequent SMART test comes up fine.
In a short testing I found that it's fine if I only do writes or reads. It only fails if it's doing both at the same time.
So, figured bad drive, no problem - swapped it with an other older drive, and it gives me the same issues.
For giggles, I figured I'd try swapping connections, just to see, and the issues follows the drive, so one would think it's not the connector (even swapped which port on the HBA was used just in case)
Two bad drives? Well they're both been in use previously so sure, certainly possible (they're completely different manufacturer and date mind you)
Just because I'm only testing anyway, I figured, let's mess around and put both of the drives that give me errors in the pool....
Here's where it gets truly odd.
I still get the same error, still only on combined reads and writes, but only on one of them at a time. They pretty much go 50/50. But no matter how hard I try, I can never fault both drives at the same time. Indeed I cannot fault the second drive even when it falls or is put offline completely.
Now, in the end, I'm just going to get a new drive and put it in at this point, but I am curious if anyone's seen this before, and if so, if they know a cause - especially if it's NOT an HDD failure I'd love to know before I bring this build 'live' (probably a week, maybe two, away still as of posting)
Oh, and I'm running latest FREENAS stable, no nightlies here.
