Description of problem: Our initiator fails to logout at the end of our session. 2.4.21-28.ELsmp
Does fail to logout mean send a logout PDU? We do not send a logout on the 4.0 version. I think it actually does on the 3.X version in certain cases though. Why do you need this? It is not required.
This was a result of connectathon. Bill, any background on why someone thought it was important?
EMC is who originally brought it up at Connectathon. I then talked about it with NetApp and I'm trying to remember what they said. They confirmed we didn't log out but said it didn't much matter to them at the moment. But Some new error handling ability in the future could be affected by this? I'll look through my notes on Monday and see if I wrote it down. Oh right.. I think I remember.. in the future, You will have to log out, just dissapearing won't log you out since they will keep the connection open for error recovery. I think the concern was that we might exhaust resources on the iSCSI target?
My inclination on this is to keep it in mind for RHEL 5, and maybe also address it in RHEL 4 if the risk and effort are low. I would not change RHEL 3 at this point unless we get a customer complaint. Mike, I'll assign it to you. Maybe take a look at open-iscsi and RHEL 4 to see if we can be better citizens by logging out when there is a controlled end of session event. You can close this BZ, or assign it to RHEL 4 or 5.
I am closing this BZ. It is not a bug. The RFC is not going to change anytime soon becuase the behavior is needed for things like erl=0. For open-iscsi we are going to do it when we get time. For RHEL3 we can log out sometimes, but that code actually has a bug which no body has seemed to notice (I just happened to find it when I was cleaniing up the RHEL4 codde :)). For RHEL4, we actually had it but took it out due to some werird interop/RFC interpretation problem or worries.