I decided to test the "ZIL on SSD" thing. I placed a 2GB partition on my GPT formatted (gpart -a 4k) 60GB Corsair Force 3 SSD. Did these then rebooted:
My root is on zfs, so the regular command above results in an error:
The work-around is this, before adding the log device:
After the log device has been added you can re-set the value:
System comes back up, but I get a strange ashift result:
Details:
1. I only have 1 zpool
2. I also did the process by booting from a USB installation of my kernel/world and did a zpool export mypool before reboot - the result was the same
3. The zdb command stays busy after displaying the message and I have to <ctrl-c> out of it.
4. I know that the log should be given the entire device, but that is so that the SSD and the on-board controller do not get drowned by I/O stream. In my case this is not an issue so partitioned log is useable.
What do you think the multiple ashifts mean? I assume it's because of my little "workaround".
# gnop create -S 4K ada1p9# zpool add mypool log ada1p9.nopMy root is on zfs, so the regular command above results in an error:
Code:
cannot add to 'mypool': root pool can not have multiple vdevs or separate logs# zpool set bootfs="" mypoolAfter the log device has been added you can re-set the value:
# zpool set bootfs=mypool mypoolSystem comes back up, but I get a strange ashift result:
# zdb mypool | grep ashift
Code:
ashift: 9
ashift: 12
ashift: 9
ashift: 12Details:
1. I only have 1 zpool
2. I also did the process by booting from a USB installation of my kernel/world and did a zpool export mypool before reboot - the result was the same
3. The zdb command stays busy after displaying the message and I have to <ctrl-c> out of it.
4. I know that the log should be given the entire device, but that is so that the SSD and the on-board controller do not get drowned by I/O stream. In my case this is not an issue so partitioned log is useable.
What do you think the multiple ashifts mean? I assume it's because of my little "workaround".
