[mythtv-users] Recordings not working
glenb at glenb.net
Wed May 26 00:12:12 UTC 2021
On Tue, 2021-05-25 at 21:24 +0100, John Pilkington wrote:
> On 25/05/2021 18:57, Bill Meek wrote:
> > On 5/25/21 12:30 PM, Gary Buhrmaster wrote:
> > > On Tue, May 25, 2021 at 5:01 PM Paul Harrison
> > > <mythtv at mythqml.net> wrote:
> > >
> > > > mythzmserver doesn't use Qt so connects to the ZM DB through
> > > > the client
> > > > library. It does set the reconnect flag but for reliability it
> > > > still
> > > > periodically kicks the DB by sending a dummy "SELECT NULL;"
> > > > query.
> > >
> > > As I recall, old (no longer supported?) versions
> > > of mysql automatically reconnected, and then
> > > it was changed to not auto reconnect to avoid
> > > resetting transaction state out from under the
> > > applications without notice, and the option was
> > > made available to revert to the previous behavior.
> > > Belt and suspenders in MythTV was almost
> > > certainly the best (only?) choice during those
> > > various transitions of mysql versions.
> > >
> > > > It's very possible that kicking the DB is not
> > > > required anymore with newer versions of the
> > > > client library but I've not tried it.
> > >
> > > If the code works, the old adage of not
> > > "improving" it often applies.
> > Just pushed a fix for master. I'll let it soak for some time
> > before back porting to v31.0.
> > Users that have the wait_timeout timeout work around should remove
> > it,
> > or you can set it to a few minutes to have the disconnect happen
> > more
> > often. mythbackend --seterbose general:debug will allow a new log
> > message: ...mythdbcon.cpp:947 (lostConnectionCheck) - SQL Native
> > Error Code: n.
> > The code will be 4031 for 20.04 users with mysql* 8.0.25 or 2006 if
> > that fix isn't in yet.
> > It's only slightly improved (in my opinion about the slightly and
> > or improved parts).
> > Looking at git log -p mythtv/libs/libmythbase/mythdbcon.cpp,
> > I believe if reconnecting is successful, the query is retried.
> > Several changes to improve this were done log ago.
> Perhaps I ought to report that my two boxes running master with
> have not had this problem; I guess that using EIT has kept the DBs
> MariaDB versions 10.4.19 Fedora 33, 5.5.68 (!!) el7
> mythtv-users mailing list
> mythtv-users at mythtv.org
> MythTV Forums: http://email.mg.glenb.net/c/eJxFjDkKwzAQAF9jdRG2vKujUJEm_5B2V3bAR5DlQH4fQ4rANMPAcKSSGEE9owU7iHcFaXDMGW1hPwYDPlnvDaEjSpl40JRWWTrop0W2rDdpao4INpBLYkcySGQLSIAeMgo7DJLUEufWXkc33jvzuCh7PVe9ftrc3nqvk6rxJ7fzkHpc-3_7AhJwNHY
same here running 31/fixes on arch. using mariadb 10.5.10 and mariadb-
clients 10.5.10. no problems seen.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the mythtv-users