Hello NSO Team,
Admin guide says:
Replication is done on a per-transaction basis to all the slaves in parallel. It can be configured to be done asynchronously (best performance) or synchronously
Where do we configure that?
best regards
Fatih
Solved! Go to Solution.
Faith
> Admin guide says:
Replication is done on a per-transaction basis to all the slaves in parallel. It can be configured to be done asynchronously (best performance) or synchronously
> Where do we configure that?
See below from man ncs.conf. Don't change this setting unless you know exactly what you're doing.
Best Regards,
/jan
/ncs-config/cdb/replication (async | sync) [sync]
When CDB replication is enabled (which it is when high-availability mode is enabled, see
/ncs-config/ha) the CDB configuration stores can be replicated either asynchronously or
synchronously. With asynchronous replication, a transaction updating the configuration is
allowed to complete as soon as the updates have been sent to the connected slaves. With
the default synchronous replication, the transaction is suspended until the updates have
been completely propagated to the slaves, and the subscribers on the slaves (if any) have
acknowledged their subscription notifications
Faith
> Admin guide says:
Replication is done on a per-transaction basis to all the slaves in parallel. It can be configured to be done asynchronously (best performance) or synchronously
> Where do we configure that?
See below from man ncs.conf. Don't change this setting unless you know exactly what you're doing.
Best Regards,
/jan
/ncs-config/cdb/replication (async | sync) [sync]
When CDB replication is enabled (which it is when high-availability mode is enabled, see
/ncs-config/ha) the CDB configuration stores can be replicated either asynchronously or
synchronously. With asynchronous replication, a transaction updating the configuration is
allowed to complete as soon as the updates have been sent to the connected slaves. With
the default synchronous replication, the transaction is suspended until the updates have
been completely propagated to the slaves, and the subscribers on the slaves (if any) have
acknowledged their subscription notifications
Hi Jan,
Thanks for this. Is there some further information to understand the gains and risks of setting/changing that?
Best regards
Fatih
Fatih,
Thanks for this. Is there some further information to understand the gains and risks of setting/changing that?
The normal, safe and sound thing to do is to run in sync mode.
The async mode is only there because of special cases, and require special design. Come talk to us (me) if you think you have a use case in this category. An example might be that you have many standby nodes, for example. With synchronous mode you'll always be waiting for the slowest standby to respond (or timeout). Async mode removes that slowness, but this is unsafe from HA perspective unless special design measures are also taken.
Best Regards,
/jan
hi Jan,
Thanks for the example that gave a bit more clarity. At the moment, our aim is to understand the options at our best so as to guide customer.
And the recommendation in your message is clear.
thanks
Fatih