From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alexander Ahring Oder Aring Date: Wed, 21 Apr 2021 12:21:15 -0400 Subject: [Cluster-devel] [PATCHv3 dlm/next 7/8] fs: dlm: add reliable connection if reconnect In-Reply-To: References: <20210326173337.44231-1-aahringo@redhat.com> <20210326173337.44231-8-aahringo@redhat.com> <20210402205351.GA24027@linux-2.home> <20210409203221.GB30244@linux-2.home> Message-ID: List-Id: To: cluster-devel.redhat.com MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Hi, On Mon, Apr 12, 2021 at 11:21 AM Alexander Ahring Oder Aring wrote: > > Hi, > > On Fri, Apr 9, 2021 at 4:32 PM Guillaume Nault wrote: > ... > > > > > > Yes, there is one thing missing. I detected a synchronization issue > > > because the node membership of a lockspace is done by a different > > > protocol handled in user space. This protocol declares the actual peer > > > of the DLM protocol. In short, there are two protocols here, whereas > > > one handles the membership of peers and the other handles the actual > > > operations. There is a missing synchronization between operations and > > > membership which I solved with a DLM_FIN message. > > > > Thanks, I'll re-read the parts about DLM_FIN with this explanation in > > mind. > > > > okay. Thank you. > Additional note: The DLM_FIN is like a 4 way close handshake with the possibility of having a half-close socket. We probably can do that on TCP with shutdown, but not on SCTP. However it's now on the application layer. - Alex