Dangit, the #FreeBSD installer foiling me again. Trying to install onto a zfs root mirror, choosing "Force 4k Sectors" (the default) says "gpart: Invalid alignment param: Invalid argument". Turning _off_ forced 4k sectors says "newfs_msdos: Invalid bytes/sector (520): must be 512, 1024, 2048, or 4096".
Post
Remote status
Replies
11@AFresh1 are you using disks with 520-byte sectors? you may want to reformat them as 512-byte, since it has no benefit with zfs
@lw maybe. I mean, this is the installer creating the partitions, I wonder how I change that.
@AFresh1 i don't know how to do it on freebsd off hand, maybe something with camcontrol. linux apparently has a utility called sg_format which can do it...
@lw Looks like `camcontrol` We'll see how it works.
@lw yay! the camcontrol in that article did fix it. Who needs to understand stuff
โ
@AFresh1 we should probably have an easier way to do this, you don't often need it but it's rather unhelpful to require the user to know the SCSI command as a sequence of bytes...
Now I got it partitioned, but screwed up the networking. Now it won't let me re-partition on the next go-around.
โ
Making progress. Got my main zpool imported and trying to reproduce what I tested working to move VMs to vm-bhyve and instead:
```
GEOM_PART: integrity check failed (zvol/ztank/bhyve/stats/disk0p4, BSD)
```
Is there a nice recommended #FreeBSD daily / weekly / monthly job with things that TrueNAS did? Like regular smart short tests and zfs scrubs?
@AFresh1 wrt smart, actually I did this a few years ago on a customer project if nobody else knows better that's probably sufficient. I'm not up on smartmon tooling, YMMV. I just wait for the zpool scrubs to get angry.
https://gist.skunkwerks.at/dch/2c709cffb61b4e8ba1ed0bf17dfe5af4
ZFS will notice when something is up with a couple checksum errors or kicking it out of the pool. You might be able to put it back and it may very well scrub clean.
The kernel will definitely notice with CAM error / timeout / retry messages but by then you are probably already triaging the issue
IMHO, the only way to detect a failing disk early is to watch the IO latency. Once a disk falls out of the normal performance range it's time to send it to the farm upstate ๐