All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Anyone with experience with Influxdb failover?
If anyone doesn't know InfluxDB removed all cluster functionality from the open source version and made it an enterprise product. https://portal.influxdata.com/
Personally I think the # core limitation is completely unreasonable, $400/month for just 2 cores!
Anyway what do you think is the best way to create some kind of redundancy? Does anyone know if simply running drbd on the influxdb directories would work?
So if master fails I would just do "service influxdb start" on the slave and everything would be exactly as when the master went down. Am I mistaken?
All I want is a fast and reliable way to start a new DB server within minutes without any data loss in case the DB server burns down.
Please dont include suggestions of alternative databases. This topic is only about influxdb and related tools/services. Thanks
Comments
The short-term solution is to not upgrade to the newer version. If you wait long enough, someone will fork the open-source version and figure out the cluster functionality. I assume this impacts a large user-base.
Blindly applying file-system replication can lead to DB corruption. If you really want to do this:
Funnel all writes to a single master node. Changes can be replicated to slaves - with DRBD/rsync whatever. Reads can go anywhere. You would get strong write consistency, and eventual consistency for reads. Failover of your master would need election of a new master. The closest similar example I can think of is RQLite which does Raft consensus on top of SQLite.
Googled and just found this: https://influxdata.com/blog/update-on-influxdb-clustering-high-availability-and-monetization/
They do have InfluxDB Relay open-sourced for handling HA.
The Relay is just a proxy that allows me to write data to multiple DBs at the same time. It does nothing to continuous queries and other stuff that run on the DB.
What I want is 1:1 copy of the whole thing. That is only possible with their enterprise cluster.
So Master-Slave would work fine with DRBD/glusterfs/rsync? Failover to master can be done manually by me. I dont need that much automation.
I cant find much information on this kind of setup.
InfluxDB has backup and restore and seems to have an incremental option. So if you just backup to multiple locations at regular intervals, that would be the safest option.
In a master/slave setup with filesystem replication (rsync/drbd), the worst that can happen is the master dies halfway through a transaction and the slave deletes the partial transaction on waking up. But that should be Ok.
If you are doing manual failover, then start the Influx server on a slave only when you need it - don't want idle slave influx daemons messing with those backups. Essentially your situation seems to require just backup with slave daemons configured to run straight from the backup.