All of lore.kernel.org
 help / color / mirror / Atom feed
* New user of tgtd
@ 2023-07-14 15:07 Forza
  2023-07-14 15:46 ` ronnie sahlberg
  0 siblings, 1 reply; 5+ messages in thread
From: Forza @ 2023-07-14 15:07 UTC (permalink / raw)
  To: stgt

Hi,

I am a new user of tgtd. I am intending to replace LIO with tgtd because it's a user-space target.

My use-case is to export files to remote initiators, so I will be using the backing-store option.

I have been trying to go through the docs and examples but I have some questions I'd like to clarify.

* aio driver. Does this mean tgtd won't pass down flush/barrier requests from initiators (like in LIO)? I. E. data loss on power failure?
* I am using btrfs on the target server and I'd like to use the thin_provisioning option. Am I correct in that this has to be enabled via "params  thin_provisioning=1" in the backing-store section?
* Is there a way to hint to the initiators that the LUN is non-rotational? Some OSes use this parameter to adjusts its internal io scheduling.
* i notice that tgtd uses 16 IO threads no matter the number of cpus. Is there a guideline on a suitable number for my use-case?
* Is ErrorRecoveryLevel=1/2 implemented in tgtd so that supporting clients transparently can recreate the connection after interruption or target restarts?
* the targets.conf.5.xml doesn't seem to include all available options. I am glad for the examples which helped me get going :)

I also noticed the following warning:

tgtd: bs_init_signalfd(267) could not open backing-store module directory /usr/lib/tgt/backing-store

I created the dir, but can't see it is actualky being used. What is the purpose of this directory?

Thanks,

~ Forza

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: New user of tgtd
  2023-07-14 15:07 New user of tgtd Forza
@ 2023-07-14 15:46 ` ronnie sahlberg
  2023-07-14 17:32   ` Forza
  0 siblings, 1 reply; 5+ messages in thread
From: ronnie sahlberg @ 2023-07-14 15:46 UTC (permalink / raw)
  To: Forza; +Cc: stgt

> * Is there a way to hint to the initiators that the LUN is non-rotational? Some OSes use this parameter to adjusts its internal io scheduling.

I think the param rotation_rate=0 will do that.

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: New user of tgtd
  2023-07-14 15:46 ` ronnie sahlberg
@ 2023-07-14 17:32   ` Forza
  2023-07-14 21:21     ` ronnie sahlberg
  0 siblings, 1 reply; 5+ messages in thread
From: Forza @ 2023-07-14 17:32 UTC (permalink / raw)
  To: ronnie sahlberg; +Cc: stgt



---- From: ronnie sahlberg <ronniesahlberg@gmail.com> -- Sent: 2023-07-14 - 17:46 ----

>> * Is there a way to hint to the initiators that the LUN is non-rotational? Some OSes use this parameter to adjusts its internal io scheduling.
> 
> I think the param rotation_rate=0 will do that.

Thank you. I didn't see it as a documented opinion, but looking through the source code[*] it seems to be the correct option.

[*] https://github.com/fujita/tgt/blob/1a5e14e678ba944986428f4abc586a68e7cdf31a/usr/tgtd.h#L93

~ Forza


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: New user of tgtd
  2023-07-14 17:32   ` Forza
@ 2023-07-14 21:21     ` ronnie sahlberg
  2023-07-15 23:01       ` Forza
  0 siblings, 1 reply; 5+ messages in thread
From: ronnie sahlberg @ 2023-07-14 21:21 UTC (permalink / raw)
  To: Forza; +Cc: stgt

On Sat, 15 Jul 2023 at 03:32, Forza <forza@tnonline.net> wrote:
>
>
>
> ---- From: ronnie sahlberg <ronniesahlberg@gmail.com> -- Sent: 2023-07-14 - 17:46 ----
>
> >> * Is there a way to hint to the initiators that the LUN is non-rotational? Some OSes use this parameter to adjusts its internal io scheduling.
> >
> > I think the param rotation_rate=0 will do that.
>
> Thank you. I didn't see it as a documented opinion, but looking through the source code[*] it seems to be the correct option.
>
> [*] https://github.com/fujita/tgt/blob/1a5e14e678ba944986428f4abc586a68e7cdf31a/usr/tgtd.h#L93

The documentation might be missing here because it is possibly an
unusual thing to need.
Maybe create a pull request to add this to the docs?


>
> ~ Forza
>

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: New user of tgtd
  2023-07-14 21:21     ` ronnie sahlberg
@ 2023-07-15 23:01       ` Forza
  0 siblings, 0 replies; 5+ messages in thread
From: Forza @ 2023-07-15 23:01 UTC (permalink / raw)
  To: ronnie sahlberg; +Cc: stgt



---- From: ronnie sahlberg <ronniesahlberg@gmail.com> -- Sent: 2023-07-14 - 23:21 ----

> On Sat, 15 Jul 2023 at 03:32, Forza <forza@tnonline.net> wrote:
>>
>>
>>
>> ---- From: ronnie sahlberg <ronniesahlberg@gmail.com> -- Sent: 2023-07-14 - 17:46 ----
>>
>> >> * Is there a way to hint to the initiators that the LUN is non-rotational? Some OSes use this parameter to adjusts its internal io scheduling.
>> >
>> > I think the param rotation_rate=0 will do that.
>>
>> Thank you. I didn't see it as a documented opinion, but looking through the source code[*] it seems to be the correct option.
>>
>> [*] https://github.com/fujita/tgt/blob/1a5e14e678ba944986428f4abc586a68e7cdf31a/usr/tgtd.h#L93
> 
> The documentation might be missing here because it is possibly an
> unusual thing to need.
> Maybe create a pull request to add this to the docs?
> 
> 

Yes, good idea. I did a small PR.

Thanks.

>>
>> ~ Forza
>>



^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2023-07-15 23:01 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-07-14 15:07 New user of tgtd Forza
2023-07-14 15:46 ` ronnie sahlberg
2023-07-14 17:32   ` Forza
2023-07-14 21:21     ` ronnie sahlberg
2023-07-15 23:01       ` Forza

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.