maria-discuss team mailing list archive

[Thorvald Hallvardsson <thorvald.hallvardsson@xxxxxxxxx>]

Hi Guillaume,

Thank you for that. The only options I have in the config file is:
router_options=max_sescmd_history=4096
That's all.

I can add it but that would require maxscale to restart would it not ?

If I set wsrep_desync=1 on one of the nodes... how do I know that node will
be marked as a donor ?

At the moment I have this in Galera config:
wsrep_cluster_address="gcomm://10.44.167.11
,10.44.167.12,10.44.167.13,10.44.167.19,10.44.167.20"

So I need to expand that list by the 2 other servers is that correct ?
That also requires a mysql restart on each node in the cluster ?

I'm just a bit worried about those restarts as we had situations in the
past when the nodes got stuck on "joining" without actually transferring
anything but I think we have actually fixed it now.

Thank you.

Regards,
TH

On 4 July 2016 at 10:29, Guillaume Lefranc <guillaume.lefranc@xxxxxxxxxxx>
wrote:

> Hi,
>
> SET GLOBAL wsrep_desync=1 will set a node as desynced and will cause
> MaxScale to stop sending traffic to it (provided you use
> router_options=synced in your maxscale.cnf config file)
>
> -GL
>
> On Mon, Jul 4, 2016 at 11:25 AM Thorvald Hallvardsson <
> thorvald.hallvardsson@xxxxxxxxx> wrote:
>
>> Hi guys,
>>
>> I need some help as I'm preparing to expand my existing Galera cluster.
>> Generally the purpose of it is I need to replace 3 VMs I have a cluster
>> running on to 2 physical servers I have been given for that purpose.
>>
>> My idea was to join 2 physical cluster to the existing cluster of 3 and
>> then remove all 3 VMs from the cluster leaving 2 physical nodes within. I
>> have a arbitrator box within the cluster as well.
>>
>> The config at the moment contains 2 maxscale proxy nodes on the front
>> balancing the read queries across the entire cluster of 3 and write queries
>> to the elected master.
>>
>> What I want to achieve is as I said above but + I will need another help
>> with maxscale to expand and then shrink the cluster down to 2 nodes
>> afterwards. I have had a situation before when I had to do a simple config
>> change on the Galera cluster and I was trying to mark nodes and put them
>> into maintenance mode to stop traffic coming to them but the commands like
>> set server NODE master didn't actually do anything. I had situations when I
>> had 2 masters in the cluster for a second because then something was
>> happening that my manually selected master was becoming a slave again. I
>> couldn't find anything on the internet what would point me into right
>> direction and explain the whole procedure in details.
>>
>> My cluster is quite busy at the moment even overnight so there is no
>> perfect time to do anything on it. I need to do it the least disruptive way
>> and the most controlled way as well. Hence I'm asking for help and
>> suggestions as I found both Galera and maxscale documentation very very
>> brief not actually providing any decent details or procedures. I found much
>> more help from Percona but descriptions don't match sometimes.
>>
>> Please help.
>>
>> Regards,
>> TH
>> _______________________________________________
>> Mailing list: https://launchpad.net/~maria-discuss
>> Post to     : maria-discuss@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~maria-discuss
>> More help   : https://help.launchpad.net/ListHelp
>>
> --
> Guillaume Lefranc
> Remote DBA Services Manager
> MariaDB Corporation
>