DPE-5178 Adopt admin-address throught out #516
Conversation
| ) | ||
| if conn.is_connected(): | ||
| connections.append(conn) | ||
| sleep(0.5) |
There was a problem hiding this comment.
sleep added to slow down tests so I could validate connections in database. Left here for the same reason, since the test is fairly quick
There was a problem hiding this comment.
It smells to me, it is a workaround for https://warthogs.atlassian.net/browse/DPE-5340 (we should sent peer details when we ready to accept traffic only).
There was a problem hiding this comment.
Not a workaround for that issue. Just to be able to observe the test while developing.
The error on the nightly tests seems a different issue to me, since the router is connecting to a standby cluster (and we have yet to change/discuss if the router should connect to database through router to pick leadership changes on async cases)
paulomach
requested review from
taurus-forever,
shayancanonical and
carlcsaposs-canonical
and removed request for
shayancanonical
| create_db_connections(1, **credentials) | ||
|
|
||
| logger.info("Get cluster status while connections are saturated") | ||
| _ = await run_action(mysql_unit, "get-cluster-status") |
There was a problem hiding this comment.
should we also add a check that the admin port is unaffected? or at least admin connections are active?
There was a problem hiding this comment.
Getting the cluster status rely on the admin connection. So when client connections are saturated, the action passing does just that. If the action fails, run_action will raise an assertion error
| ) | ||
| if conn.is_connected(): | ||
| connections.append(conn) | ||
| sleep(0.5) |
There was a problem hiding this comment.
It smells to me, it is a workaround for https://warthogs.atlassian.net/browse/DPE-5340 (we should sent peer details when we ready to accept traffic only).
|
* on ip change, restart mysqld * ip change custom charm event needs to be defined on charm object
Issue
As part of cycle, we ought to utilize admin-address for charm usage.
The idea is to ensure that the charm can connect to mysqld
instances (local or remote unit) even when
max_connectionsis saturatedby related application(s), since connections to admin-address are do not count against
max_connections.Solution
admin_addressin configuration file.admin_address:admin_portfor users withADMIN_CONNECTIONprivilege.s/clusteradmin/serverconfigdue cluster admin not having (and not needing to have, since replication channels don't count againstmax_connection)ADMIN_CONNECTIONprivilege.