Changeset View
Changeset View
Standalone View
Standalone View
src/docs/user/cluster/cluster_databases.diviner
Show First 20 Lines • Show All 295 Lines • ▼ Show 20 Lines | |||||
be slow, so offloading it to a replica can make the perforance of the master | be slow, so offloading it to a replica can make the perforance of the master | ||||
more consistent. | more consistent. | ||||
To dump from a replica, wait for this TODO to be resolved and then do whatever | To dump from a replica, wait for this TODO to be resolved and then do whatever | ||||
it says to do: | it says to do: | ||||
TODO: Make `bin/storage dump` replica-aware. See T10758. | TODO: Make `bin/storage dump` replica-aware. See T10758. | ||||
With recent versions of MySQL, it is also possible to configure a //delayed// | |||||
replica which intentionally lags behind the master (say, by 12 hours). In the | |||||
event of a bad mutation, this could give you a larger window of time to | |||||
recognize the issue and recover the lost data from the delayed replica (which | |||||
might be quick) without needing to restore backups (which might be very slow). | |||||
Delayed replication is outside the scope of this document, but may be worth | |||||
considering as an additional data security step on top of backup snapshots | |||||
depending on your resources and needs. If you configure a delayed replica, do | |||||
not add it to the `cluster.databases` configuration: Phabricator should never | |||||
send traffic to it, and does not need to know about it. | |||||
Next Steps | Next Steps | ||||
========== | ========== | ||||
Continue by: | Continue by: | ||||
- returning to @{article:Clustering Introduction}. | - returning to @{article:Clustering Introduction}. |