> > here is some of what one of the iscsi dev guys at
> ibm
> > said:
> >
> > It appears you are correct about the MaxCmdSN and
> > ExpCmdSN. If they are equal, that should mean we
> have
> > a command queue depth of one. But it looks like
> our
> > iSCSI driver has an off-by-one-error when
> determining
> > if the command queue is full. I assume that after
> the
> > login, there is a delay of a minute or so before
> the
> > connection is closed and restarted. Is that the
> case?
>
> yes, that is the case.
>
> > Do you actually have a target that wants to use a
> > command queue depth of just 1? Most have anywhere
> from
> > 16-128 or more and so we had not hit this before.
> BTW,
> > what iSCSI target are you using?
>
> we did not adjust that until FFP. now we adjust that
> to 2 before FFP.
>
> >
> > As for the configuration, we curently do not have
> > definite plans for extending it, though I have
> some
> > ideas of what I would like to add and I have had a
> few
> > (but not many) other customers make similar
> comments.
> > I will add it to our list of enhancements to
> consider
> > but don't expect any short-term changes.
> >
> > We don't have any particular tools that we
> recommend.
> > Some of our stuff is for internal use only, though
> > there is at least one publicly available tool that
> we
> > use for file system stress....a program called
> "dt".
> > One place you can get it is from the website for
> > "connectathon"....www.connectathon.org. Find the
> test
> > suite for iSCSI and it contains "dt".
>
> i see. i have dt.
ibm is working on the speed issue with iet--aix. they
asked me to reset the target for aix to low settings
and in syslog on iet machine i see:
check_segment_length(1552) too lond data 53 65536 8192
i know what it means however i think the wording could
be changed a bit. here is a patch to fix a "typo"...
unless of course this message was meant to be as it
was...
--- kernel/iscsi.c.old 2006-01-17 12:24:47.526973468
-0500
+++ kernel/iscsi.c 2006-01-17 12:25:41.954606342
-0500
@@ -1548,7 +1548,7 @@
struct iscsi_sess_param *param =
&conn->session->param;
if (cmnd->pdu.datasize >
param->max_recv_data_length) {
- eprintk("too lond data %x %u %u\n",
cmnd_itt(cmnd),
+ eprintk("data too long%x %u %u\n",
cmnd_itt(cmnd),
cmnd->pdu.datasize,
param->max_recv_data_length);
if (get_pgcnt(cmnd->pdu.datasize, 0) >
ISCSI_CONN_IOV_MAX) {
aaarrrggghhh!!!!
FreeBSD rocks
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems? Stop! Download the new AJAX search engine that makes
searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=103432&bid=230486&dat=121642
|
