Web DB

The web database is a Postgres DB running on port 5432. If you have multiple nodes with this role, then the first node in the list is the primary and the others are streaming replicas.

DB troubleshooting can be performed by an experienced Database Administrator (DBA). For help in this matter, please contact Circonus Support (support@circonus.com).

At times, it is necessary to log into the database to run requested queries. To log in from the local machine, issue the following command:

/opt/pgsql/bin/psql -U circduown -d circonus

This will log you in as a non super user, but with the capability of running the appropriate queries.

The Postgres data directory is located at /wdb/pgdata/9.2

In this directory there is also a pg_log directly, which will contain log files useful for general debugging of database connectivity issues. It also contains some queries.

Note that any config files located in the data directory are managed by Circonus. Any manual changes will not persist if run-hooper is used. Some common tuning parameters can be changed via site.json. If you need additional customizations beyond what Hooper provides, please contact Circonus Support (support@circonus.com).

Web DB Failover

In the event of a database failure, it will be necessary to manually failover to one of your replicas, which will become the new primary. To do this, use the following steps.

  1. If the current primary is still running, stop the circonus-postgres-circonus_wdb service.
  2. On the new primary, make a copy of the file /wdb/pgdata/9.2/recovery.conf to a location outside of the /wdb/pgdata hierarchy, such as your home directory.
  3. If using an IP alias as the connect_host, relocate the IP/DNS name to the new primary. Otherwise, update site.json, setting a new hostname for master and connect_host in the web_db object.
  4. On the new primary, touch the file /wdb/pgdata/9.2/failover.now.
    • When the new primary completes its failover, the file /wdb/pgdata/9.2/recovery.conf will be renamed to “recovery.done”.
    • After the renaming occurs, you can delete the failover.now file, if it is still present.
  5. Run Hooper on the new primary to ensure any configuration customizations are applied: /opt/circonus/bin/run-hooper -m, restarting the database if directed. Hooper will show the names and locations of services that must be manually restarted after a primary database restart. Complete those restarts before continuing.
  6. Distribute the updated site.json to all other hosts in your deployment, and perform a Hooper run (/opt/circonus/bin/run-hooper -m) across all hosts, following the order specified in the initial installation page.
  7. At this point, your deployment should be failed over to the new primary.
  8. Rebuild any other replicas off of the new primary. Follow these steps on each machine:
    1. Stop the circonus-postgres-circonus_wdb service.
    2. Run the command: rm -rf /wdb/pgdata/9.2
    3. Run the command: mkdir /wdb/pgdata/9.2
    4. Run the command: chown postgres:postgres /wdb/pgdata/9.2
    5. Run the command: chmod 700 /wdb/pgdata/9.2
    6. Run the command: sudo -u postgres /opt/pgsql/bin/pg_basebackup -D /wdb/pgdata/9.2/ -h <new primary host/IP> -U replication
    7. Edit the recovery.conf file from Step 2, changing the host in the primary_conninfo parameter to point to the new primary.
    8. Place the recovery.conf into the path /wdb/pgdata/9.2/recovery.conf, overwriting the current file in that path.
    9. Start the circonus-postgres-circonus_wdb service.
  9. If/when the problems with the old primary are addressed, make it into a replica by running the rebuild commands from the previous step on that machine.

Web DB Restart

If certain configuration parameters are modified, Hooper will notify the operator that a database restart is required. Due to the potential disruption to related services, database restarts are not done automatically. If you need to restart the Web DB, execute one of the following commands, depending on the OS:

  • RHEL/CentOS: systemctl restart circonus-postgres-circonus_wdb

Then restart each of the services that we recommend restarting after a Web DB restart. See Service Dependencies.