Hi Vic,
On 9/28/06, Vic Engle <[EMAIL PROTECTED]> wrote:
Thanks for the info. After having read about the tunable parameters
my impression was that the default values were designed to work in
most environments but in some specific environments they might not
be optimal.
Yes, that's the idea. Sun has put a lot of effort into making that happen.
My thinking with the ssd_max_throttle was that path failover
time would be a little quicker if my host has to retry fewer commands
so I was thinking that setting max_throttle to as low a value as possible
without affecting I/O performance on my host would be desirable. Also,
this would provide some protection to me against having a bunch of
my hosts fill up the target port queue on my array.
I'm not sure that you're understanding this correctly. The sd_max_throttle
variable is the upper limit on the number of outstanding commands in the
queue. This is distinct from retries. For various arrays there are vendor-
sourced recommendations on what to set this variable to.
Be aware, however, that sd_max_throttle is a *global* tunable, and not
something which you can set on a per-lun, per-target or per-hba basis.
The fp_offline_ticker, according to a doc I read about this stack, has
a default value of 90 seconds. That seems a bit long for a small to
medium 2Gb fabric, so I thought I might improve path failover here again.
As I said, wait for a Solaris 10 update. I believe that this change (which
I integrated) is due for Solaris 10 update 3, which is expected to be
released as 11/06.
I don't think you should be looking at those defaults as being too long
or too short for a fabric which operates at a certain speed. Unless you
have specific concerns which are backed up by empirical data then it
should be unnecessary to change.
The ssd_io_parameter is 60 seconds by default. I don't think I have
very many normal situations where I would reach this limit so I would
just prefer to retry/failover sooner.
Again, if you don't have empirical data which proves that this setting
is an issue in your environment, don't change it.
I'm not certain I want to adjust the ssd_retry_count but I would like to
be certain of the default behavior and since I get the error when I have
ssd_retry_count in my system file I was concerned about what that
parameter was really set to.
The default behaviour is fine for the vast majority of installations. You
should not need to change it. If you are getting errors with your upper
layers (eg, Oracle) then a Sun support engineer will make a suggestion
about what to do to make the value more appropriate.
In your reply, were you saying that I could use sd_* parameters instead
of ssd_* parameters to adjust ssd throttles on sparc systems?
Yes, that is correct.
James C. McPherson
--
Solaris kernel software engineer, system admin and troubleshooter
http://www.jmcp.homeunix.com/blog
Find me on LinkedIn @ http://www.linkedin.com/in/jamescmcpherson
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss
