Without MA, HA does not work

Help
1

Hi,

Take a look at the following scenario. I hope the HA of SingStore DB is improved. (Is there any good way?)

  1. The cluster is deployed on two hosts as follows:
    host#1 (MA + LF1), host#2 (CA + LF2)

  2. Enable HA.

  3. Force host#1 OFFLINE.
    (I couldn’t actually shutdown, so I excuted the sdb-admin stop-node command in the order MA, LF1.)

  4. Then, SingleStore DB is OFFLINE.

  5. Connect to the CA and run “AGGEGATOR AS SET MASTER;”.
    Now, SingleStore DB will be ONLINE.

Here,
If step 5 is not performed, SingleStore DB cannot be ONLINE. I hope there was a way to do this automatically.

Thank you.

2

Hi,

No we don’t have automatic failover for the master aggregator today. It’s on the roadmap for a future release. Its best practice to have the MA on its own host because of this. You only need the MA to coordinate failovers of other nodes and for DDL, so if only the MA is down things will stay online.

-Adam

1 Like