On 03/28/2011 11:13 AM, Or Gerlitz wrote:
On 3/27/2011 7:46 PM, Stepan Fedorov wrote:
We use SLES. May be we can backport this fix to SLES 2.6.32 kernel?
I can look on that, any idea/pointer for public download of the src rpm
for their kernel? Also, I wasn't sure who's the SLES maintainer for
We use SLES. May be we can backport this fix to SLES 2.6.32 kernel?
27.03.2011, в 12:24, Or Gerlitz написал(а):
Mike Christie wrote:
ccing the iSER maintianer.
Off the top of my head it looks like we could not allocate something then
when handling the error we might be trying to use the
Hello,
Recently we have started to test our iscsi targets using Ubuntu hosts
with open-iscsi initiator. Our tests use sg to make IOs. We have seen
that on high load there are delays for some IOs. Sometime it takes 40
sec.
My investigation shows that the IOs are delayed somewere in the Ubuntu
Mike Christie wrote:
ccing the iSER maintianer.
Off the top of my head it looks like we could not allocate something
then when handling the error we might be trying to use the null pointer, tight?
Yep, we made a recent fix around that flow, Stepan, could you please try
with a kernel that has
upfront, there's max_fmrs but its zero...). Is that 123 targets for
specific initiator a realistic requirement for your environment?
No. Realistic requirement is at least 256 targets per node.
Each target represents virtual machine in VDS hosting environment, so
we need at least two targets
Hm... it seems like i can login only on 123 targets. Next login have
this in dmesg:
[ 6705.243573] sd 462:0:0:2: [sdij] Attached SCSI disk
[ 6811.884501] iser: iser_connect:connecting to: 192.168.15.5, port 0xbc0c
[ 6811.886415] iser: iser_cma_handler:event 0 conn 8801d6beee10 id
OK. I apply patch on SLES kernel, and test attaching of 100 targets
with two LUNs on each.
Before applying patch, i was a subject message in dmesg.
After applying patch, there is messages like this in dmesg:
[ 2719.443306] iser: iser_create_ib_conn_res:unable to alloc mem or
create resource, err
Степан Фёдоров wrote:
i can login only on 123 targets. Next login have this in dmesg:
[ 6811.893602] fmr_pool: fmr_create failed for FMR 184
[ 6811.893779] iser: iser_create_ib_conn_res:unable to alloc mem or create
resource, err -12
If next i logout one session, then i can login one target
Ooops:
Before applying patch, i was a subject message in dmesg.
* there was
But in top on two initiator nodes i see only about 2Gb of memury used
* memory
--
You received this message because you are subscribed to the Google Groups
open-iscsi group.
To post to this group, send email to
On 3/27/2011 7:46 PM, Stepan Fedorov wrote:
We use SLES. May be we can backport this fix to SLES 2.6.32 kernel?
I can look on that, any idea/pointer for public download of the src rpm
for their kernel? Also, I wasn't sure who's the SLES maintainer for
iscsi nowadays, Hannes, is that still you?
Hi Mike, I don't see this message in the open-iscsi archive, I also
didn't get it to a subscription I have in my gmail email account. I'm
suspecting this is b/c my sender address is now @mellanox.com which is
different from the email I'm subscribed with - @voltaire.com, the group
page says
On 03/27/2011 03:08 AM, Evgeny Schmeilin wrote:
Hello,
Recently we have started to test our iscsi targets using Ubuntu hosts
with open-iscsi initiator. Our tests use sg to make IOs. We have seen
that on high load there are delays for some IOs. Sometime it takes 40
sec.
My investigation shows
On 03/28/2011 03:38 AM, Matthew Richardson wrote:
Looking at the logs (attached) there seems to be quick noticing of the
connection error (cable unplug), then recovery, then a long period
(around 360 seconds), then a second error/recovery (no action on my part
to trigger this). The disk writes
On 03/28/2011 03:34 PM, Mike Christie wrote:
On 03/28/2011 03:38 AM, Matthew Richardson wrote:
Looking at the logs (attached) there seems to be quick noticing of the
connection error (cable unplug), then recovery, then a long period
When did you plug the cable back in?
(around 360
14 matches
Mail list logo
