As I remember, GET_LOCK is supported if wsrep_on is ON in latest mariadb
versions.

-Wei

On Wednesday, August 14, 2024, Joan g <joang...@gmail.com> wrote:

> As Rohit mentioned, since the GET_LOCK function isn't supported here, could
> there be a risk of inconsistency* if both managers update the same row
> simultaneously*?
>
> @João Jandre Paraquetti Thank you pointing out I was connecting to  'cloud'
> DB for db.usage.uri
>
>
> Jon
>
> On Wed, Aug 14, 2024 at 9:59 AM Rohit Yadav <rohit.ya...@shapeblue.com>
> wrote:
>
> > Worth mentioning here that due to GET_LOCK limitations you may not use
> > your db cluster in active-active configuration.
> >
> > Regards.
> >
> >
> >
> > ________________________________
> > From: Marty Godsey <mar...@rudio.net>
> > Sent: Tuesday, August 13, 2024 11:22:47 PM
> > To: users@cloudstack.apache.org <users@cloudstack.apache.org>
> > Subject: Re: Cloudstack 4.19 Behavior with MariaDB Galera Cluster
> >
> > I use a Galera Cluster for my DB, which works without issue. I do not,
> for
> > other reasons, load balance. It is set to failover. Before I had to make
> > this change, I used load balance, and it worked fine, however.
> >
> > Regards,
> > Marty Godsey
> >
> >
> > From: João Jandre Paraquetti <j...@scclouds.com.br>
> > Date: Tuesday, August 13, 2024 at 1:46 PM
> > To: users@cloudstack.apache.org <users@cloudstack.apache.org>
> > Subject: Re: Cloudstack 4.19 Behavior with MariaDB Galera Cluster
> > WARNING: This email originated from outside of the organization. Do not
> > click links or open attachments unless you recognize the sender and know
> > the content is safe.
> >
> >
> > Hello Jon,
> >
> > Whether there will be randomness or not depends on your setup. As far as
> > I know, there are two possible modes for the connection with Galera
> > clusters: sequential and loadbalance. With the sequential mode, the
> > connector will try to connect to hosts in the order in which they were
> > declared in the connection URL, so no randomness. Using the loadbalance
> > mode, the connector performs load-balancing for all queries by randomly
> > picking a host from the connection URL for each connection, so even with
> > all hosts up, you'll have random connections.
> >
> > Regarding whether random connections are a problem, since you'll be
> > using a Galera cluster, any MariaDB node should be basically the same
> > (from the mgmt servers perspective), as all data is replicated, so it
> > does not really matter which one is used. I don't see any problem with
> > different mgmt servers connecting to different MariaDB nodes that are on
> > the same Galera cluster.
> >
> > Could you share the value that you configured for db.usage.uri? please
> > make sure that you are connecting to the cloud_usage DB and not the
> > cloud DB.
> >
> > Best regards
> >
> > João Jandre
> >
> > On 8/13/24 08:24, Joan g wrote:
> > > Hello Community,
> > >
> > > I am trying to explore the mariadb flexible URI introduced in 4.19 Ref:
> > >
> > https://atpscan.global.hornetsecurity.com?d=-s6c4P8H0F-6-
> lmZZ6U00SpQrMbRGp0imtCmO17q9oA&f=SWvzu6arJFaC7g8oD58dV-
> rNUGDjTikFIHTyuWVLU7GWtnNyMz6Tg6x1Z2wjH2Zw&i=&k=Ie9V&m=jUm3E2gaksTIskF_
> WQRJEDhPUC8imjmYn5p6CgRIz4HqqTgZR46VEfxgYcZ4WOeDSSi30TuzzkpM7H-
> MjQmDfubymBkJ5ggkbm1wjD6mwd7HpdXHHnYDkyKCXoMGsSTE&n=
> OD2YhJhOuOTBvltNv5vw7yY5G1B0ApS-wcJDDiSNEKF9kUMWcZo4ux2s7IMzOwcL&r=
> CujZfpt5OxUjowOuEiunB9ROUBFRuQ2BIvaP-f6zEKC3KLuhFgDPL3jXADWhIqyQ&s=
> 100aa56155ed1380bc3f081e5584c81a68ecac1ca324e8940189c7bb1089
> 2f3a&u=https%3A%2F%2Fgithub.com%2Fapache%2Fcloudstack%2Fpull%2F7895
> > > My concern is on deployment with 2 or 3 Cloudstack management Servers.
> > >
> > > In the event of a server failure or restart, there's a possibility that
> > > each management service could connect to different MariaDB hosts. Is
> this
> > > behavior acceptable? Will it cause any issues with the database? I've
> > heard
> > > that all management servers should  connect to the same MariaDB host.
> > >
> > > Also Usage service is failing to start using  db.usage.uri, management
> is
> > > fine
> > >
> > > Jon
> > >
> >
>

Reply via email to