Thanks for the help Mike. I was able to get a wireshark trace of this problem and found that the target was misbehaving by returning out-of-order DataIn PDUs even though open-iscsi initiator had negotiated DataPDUInOrder=Yes. Turns out there was a firmware update available that fixed this problem in the iSCSI target.
On Fri, Mar 30, 2012 at 9:42 PM, Mike Christie <micha...@cs.wisc.edu> wrote: > On 03/30/2012 03:36 PM, iscsi developer man wrote: >> I'm trying to debug the following open-iscsi initiator errors that are >> reported in /var/log/messages: > > Is there something before this in the logs? > > It is rare to see this error value. I have only have seen this when the > scsi eh runs and we are sending aborts and resets and the target and > initiator get out of sync and we end up seeing a task response for a > task we thought was aborted or cleaned up from a abort. > > Trying doing > > echo 1 > /sys/modules/libiscsi/parameters/debug_libiscsi_eh > > Then rerun your targets. Do you see something about aborts and resets > getting sent before the 1010 error, > >> >> Mar 30 14:19:31 R415-08 kernel: [56802.859946] connection194:0: >> detected conn error (1010) >> Mar 30 14:19:34 R415-08 kernel: [56805.373323] connection194:0: >> detected conn error (1020) >> Mar 30 14:21:58 R415-08 kernel: [56949.620808] connection117:0: >> detected conn error (1010) >> Mar 30 14:22:01 R415-08 kernel: [56952.634359] connection117:0: >> detected conn error (1020) >> >> From the investigation I've done so far, the "1010" error means >> ISCSI_ERR_BAD_ITT, which I assume means that the initiator received an >> iSCSI PDU that contained an ITT that was unexpected. >> >> The "1020" error always follows the 1010 error and it apparently means >> the initiator closed the connection, is that correct? > > You normally see it when the target closes the connection on us, but > with some targets you can see it if we send a logout. The target might > send the logout response and close the connection really quickly and > before we completely handle the logout response we see the connection > closed event notification. > >> >> Is there a way to get the unexpected ITT so I can compare it to the >> iSCSI target debug logs and a wireshark network trace? > > Not really. It is normally a bad target type of issue, or a race/code bug. > >> >> Would setting the iscsid debug level to a non-zero value help? If so, > > No. -- You received this message because you are subscribed to the Google Groups "open-iscsi" group. To post to this group, send email to open-iscsi@googlegroups.com. To unsubscribe from this group, send email to open-iscsi+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/open-iscsi?hl=en.
