yep we think this is exactly the issue...we have been testing a new build with a later driver in UAT this morning and no recurrence of the issue so looking most probable. thanks all for the very prompt and clear responses it's most appreciated. On Tue, 17 Aug 2021 at 13:09, Dave Cramer <davecramer@postgres.rocks> wrote: > > > Dave Cramer > www.postgres.rocks > > > On Mon, 16 Aug 2021 at 12:32, Andy Hall <andyjohnhall@xxxxxxxxx> wrote: >> >> thanks for the very clear explanation much appreciated shall take this >> back to the team. >> >> On Mon, 16 Aug 2021 at 17:14, Tom Lane <tgl@xxxxxxxxxxxxx> wrote: >> > >> > Andy Hall <andyjohnhall@xxxxxxxxx> writes: >> > > we have an odd issue where a java app using JDBC which has been >> > > working fine has suddenly started disconnecting after we setup >> > > replication from the database it was connecting to ( which is now the >> > > primary in a replicating pair ) with errors such as the following... >> > >> > > CLIENT SIDE: DBNAME/logs/20210815-200601.20966.log:SessionManagerImpl >> > > WARN 20210816-00:01:00.252 - A session lost connection with the >> > > database. Disabling all sessions >> > > SERVER SIDE: 2021-08-16 01:48:27 BST FATAL: unsupported frontend >> > > protocol 1234.5680: server supports 1.0 to 3.0 >> > >> > Hmm. That "protocol version" is a GSS request: >> > >> > #define NEGOTIATE_GSS_CODE PG_PROTOCOL(1234,5680) >> > >> > So >> > >> > 1. You are using a server version that predates our GSS support. >> > >> > 2. On the other hand, the client knows GSS and is trying to use it. >> > The client should fall back to a non-GSS connection upon receiving >> > this error response. But ... >> > >> > 3. Something --- it's not very clear what --- is seeing the error >> > and going into an unwarranted panic. >> > >> > I'd try to figure out what's issuing the "Disabling all sessions" >> > message and then filing a complaint with the authors of that. > > > What version of the JDBC driver are you using? There was one version that unfortunately blindly sent out the GSS request. > That has been rectified in later versions, I believe somewhere around 42.2.20 or so. > > Although as Tom mentioned the driver does know how to deal with this, however there are some cloud providers that chose to just drop the connection > > Dave
