[Stgt-devel] Open connections
Tue Apr 8 16:54:17 CEST 2008
FUJITA Tomonori schrieb:
> On Tue, 08 Apr 2008 15:36:52 +0200
> Tomasz Chmielewski <mangoo at wpkg.org> wrote:
>> FUJITA Tomonori schrieb:
>>> On Tue, 08 Apr 2008 14:59:00 +0200
>>> Tomasz Chmielewski <mangoo at wpkg.org> wrote:
>>>> FUJITA Tomonori schrieb:
>>>>> On Sun, 30 Mar 2008 05:52:16 +0200 (CEST)
>>>>> "Niels de Carpentier" <stgt at decarpentier.com> wrote:
>>>>>>>> So it looks like the cmd_hash_list is not actually cleared on a
>>>>>>>> Is there anything more I can do to troubleshoot this issue?
>>>>>>> Can you try this patch?
>>>>>> This patch fixes the issue for me.
>>>>> Nice, thanks for testing. I've merged the patch.
>>>> There is something wrong with the "tgtadm --op show --mode target"
>>>> output - see these files:
>>> What version do you use, the git head?
>> Will this:
>> git clone git://git.kernel.org/pub/scm/linux/kernel/git/tomo/tgt.git
>> get me the git head? Because that's what I'm using.
> Yeah, you should use the git head. To make sure, try `git-log`, then
> you will see:
> commit 6a49758ea080ddc095c76d5ccce51e0c911b8e32
> Author: Doron Shoham <dorons at voltaire.com>
> Date: Mon Apr 7 16:55:21 2008 +0300
> tgt-setup-lun: fix the usage example
> Signed-off-by: Doron Shoham <dorons at voltaire.com>
> Signed-off-by: FUJITA Tomonori <fujita.tomonori at lab.ntt.co.jp>
Yep, that's it.
>> And it "feels" rather unstable - although I can boot a diskless
>> initiator (connect using iscsistart from open-iscsi), it remounts
>> read-only randomly several seconds after we start to read data from the
> Hmm, that's bad. At least, before apllying this patch, it worked
> better (though we had the open connection problem), right?
That's more problematic.
The last version I used was the one found in
http://stgt.berlios.de/releases/ - I had rather unpleasant experiences
with using development versions of other Linux iSCSI targets, so I
decided to stick with at least semi-stable tar.bz2 package.
As I had the same problem with open connections when initiator was
restarted, I though I'd give development version of tgt a try.
It mostly solves my problem, but sometimes the connection dies
(connection times out, remount ro etc.) after restarting.
Other problem I had with today's git is that tgtd "hangs", i.e., the
process still exists, but:
# tgtadm --op show --mode target
tgtadm: can't connect to the tgt daemon, Connection refused
tgtadm: can't send the request to the tgt daemon, Transport endpoint is
It happened twice or three times, but I'm not sure how to reproduce
either of the issues reliably. At least since I restarted tgtd, which
doesn't show broken display now any more.
More information about the stgt