Is it safe to use logbias=throughput on root disk?

[chrone]

Hi Proxmoxers,

I've been using ZFS mirror rpool on HDDs with ZFS ZVOL on rpool/data. Based on recommendation from open-zfs for Innodb and Postgresql volume, I found out using logbias=throughput increase the synchronous write performance significantly thanks to less double write.

Is it safe to set logbias=throughput on root disk (rpool/ROOT/pve-1)? It's not bad like using sync=disabled, right?

Reference:
http://open-zfs.org/wiki/Performance_tuning

 

[fabian]

Hi Proxmoxers,

I've been using ZFS mirror rpool on HDDs with ZFS ZVOL on rpool/data. Based on recommendation from open-zfs for Innodb and Postgresql volume, I found out using logbias=throughput increase the synchronous write performance significantly thanks to less double write.

Is it safe to set logbias=throughput on root disk (rpool/ROOT/pve-1)? It's not bad like using sync=disabled, right?

Reference:
http://open-zfs.org/wiki/Performance_tuning

logbias should not affect consistency. it might affect performance (one way or another) though, depending on the I/O workload and pool setup.

 

[chrone]

logbias should not affect consistency. it might affect performance (one way or another) though, depending on the I/O workload and pool setup.

Awesome! Thanks for the confirmation.

Enterprise SSD in my country is very expensive and we often end up using spinning disks without any SLOG devices. With logbias=throughput, the ZFS filesystem and volume synchronous write performance throughput increase a lot, almost similar to other filesystem.