[OmniOS-discuss] performance extrremely low from nothing !

Jim Klimov jimklimov at cos.ru
Tue May 24 18:39:13 UTC 2016


24 мая 2016 г. 18:41:19 CEST, "Fábio Rabelo" <fabio at fabiorabelo.wiki.br> пишет:
>Competely lost here ...
>
>Found what command replaced the "zfs_vdev_max_pending" , and it make
>no difference !
>
>First, I added this command to /etc/system :
>
>set zfs:zfs_vdev_max_active = 1
>
>make no diference, then added all this parameters :
>
>set zfs:zfs_vdev_async_read_max_active  = 3
>set zfs:zfs_vdev_async_read_min_active  = 1
>set zfs:zfs_vdev_async_write_max_active  = 11
>set zfs:zfs_vdev_async_write_min_active  = 1
>set zfs:zfs_vdev_max_active = 1
>set zfs:zfs_vdev_scrub_max_active = 2
>set zfs:zfs_vdev_scrub_min_active = 1
>set zfs:zfs_vdev_sync_read_max_active  = 10
>set zfs:zfs_vdev_sync_read_min_active  = 10
>set zfs:zfs_vdev_sync_write_max_active = 1
>set zfs:zfs_vdev_sync_write_min_active = 1
>set zfs:zfs_vdev_async_write_active_max_dirty_percent = 60
>set zfs:zfs_vdev_async_write_active_min_dirty_percent = 30
>
>no diference either ...
>
>Yes. I rebooted the machine for the first and the second try .
>
>Someone has any clue ?
>
>
>Fábio Rabelo
>_______________________________________________
>OmniOS-discuss mailing list
>OmniOS-discuss at lists.omniti.com
>http://lists.omniti.com/mailman/listinfo/omnios-discuss

How full is your pool? When it has under 10-30% (empiric value, depends on usage patterns) free space becomes so fragmented it takes longer to find a suitably sized hole and write speed collapses - often suddenly by several times or more. Does this fit?

Otherwise, what does 'iostat -Xnz' say about disks and pools?

Jim
--
Typos courtesy of K-9 Mail on my Samsung Android


More information about the OmniOS-discuss mailing list