On 2020-10-08 14:56, Phil Stracchino wrote:
Well, I understand that perspective, but this problem appeared
instantly
as soon as I upgraded to 9.6.5, with a DB configuration I've been using
for years without a hiccup. I could turn the problem on and off like a
lightswitch by updating *ONLY* the Director to 9.6.5 or rolling it back
to 9.6.3, even if connecting directly to the local DB node without
using
HAproxy. 9.6.5 Director, even without HAproxy: About one in three
jobs
hang. 9.6.3 Director, even WITH Haproxy: No hangs.
So far, 9.6.6. is behaving like 9.6.3.
Interesting.
In that case you are probably right.
Are you using MariaDB cluster because you need high availability
or because you want to achieve higher database throughput?
High availability, fundamentally. I'd honestly prefer to be using
Percona XtraDB Cluster, but there is no working, maintained ebuild for
Gentoo Linux — and enterprise customers ARE going to try to use it
against HA clusters, so we'd better be sure it works. Which it does,
as
long as attribute spooling is disabled (and you're not using Director
9.6.5).
Do you have to turn off attribute spooling with 9.6.3 and 9.6.6?
Disabling attribute spooling will inflict noticeable performance
degradation.
--
Josip Deanovic
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
