Skip to Content.
Sympa Menu

en - Re: [sympa-users] Sympa and Percona Server for MySQL

Subject: The mailing list for listmasters using Sympa

List archive

Chronological Thread  
  • From: Phil Stracchino <address@concealed>
  • To: Antoni Garcia Navarro <address@concealed>
  • Cc: address@concealed
  • Subject: Re: [sympa-users] Sympa and Percona Server for MySQL
  • Date: Wed, 15 Sep 2021 11:21:45 -0400

On 9/15/21 6:10 AM, Antoni Garcia Navarro wrote:
> Hi Phil,
>
> The problem, according to Percona documentation, is that some write
> operations cannot be propagated to the other nodes. On our
> installation, PXC Strict Mode is set to ENFORCING, and this causes the
> extra validations.
>
> Of course, I could disable validations to import the data, and reenable
> them later, but I'm guessing it could led to problems transferring data
> between nodes. Because of that, if tables could have their primary keys
> in place, Percona will be a "validated" backend for Sympa.


Making a more detailed check:

admin_table, conf_table, exclusion_table, inclusion_table, list_table,
netidmap_table, notification_table, one_time_ticket_table,
session_table, subscriber_table, user_table all have PRIMARY keys.

logs_table, stat_counter_table, stat_table do not.


Looking at logs_table, it has no obvious primary key. Adding an id
column would serve little purpose.

Perhaps (action_logs, client_logs, daemon_logs, date_logs,
error_type_logs, list_logs, msg_id_logs) is a potential PRIMARY key for
this table. It would be on the large-ish side, but it appears that it
should be unique, as Sympa should never take the same action resulting
in the same error state twice on the same message-ID at the same timestamp.


For stat_counter_table, I'm more than half tempted to just create a
covering index. There's only six columns, they're all small, and so far
all rows are unique.


For stat_table, looking at the little data I have, I think I'd go with
(daemon_stat, date_stat, email_stat, list_stat, operation_stat,
parameter_stat). But again it's so small one might almost as well just
create a covering index.


In my case there is only one robot, so the robot_stat and robot_counter
columns are pretty much moot, and I suspect stat_table.read_stat is as
well (so far it is always 1).


AUTHORS: Could you perhaps tell us what the major queries against these
three tables are? This would help in choosing an appropriately ordered
PRIMARY key.


--
Phil Stracchino
Babylon Communications
address@concealed
address@concealed
Landline: +1.603.293.8485
Mobile: +1.603.998.69



Archive powered by MHonArc 2.6.19+.

Top of Page