[zfs-discuss] Low read performance on ZFS based on iSCSI disk
kaazoo at kernelpanik.net
Wed Jan 13 05:04:21 EST 2016
Am 12.01.16 um 16:58 schrieb Chris Siebenmann:
> Normally ZFS readahead and perhaps IO aggregation will hide any delays
> due to this from you. If there is a performance hit here, it sort of
> suggests that ZFS is not able to issue all that many concurrent IOs to
> the 'device' (here the iSCSI target) at once. You may want to look into
> this, although I don't know how to see eg the queue depth that iSCSI
> initiator is advertising and supports.
on the iSCSI initiator side I changed 'node.session.cmds_max' from 1024
to 2048 and 'node.session.queue_depth' from 128 to 1024. These are the
maximum allowed values.
However, on the iSCSI target side I can only set the 'queue_depth' to a
maximum of 128, because 'hw_queue_depth' seems to be detected as 128.
I also changed 'default_cmdsn_depth' from 128 to 512.
The read performance didn't improve after applying these changes and
re-discovering and re-connecting to the target.
More information about the zfs-discuss