When defining and running replication processes, follow these
recommendations.
- Before replicating to production, run each replication from staging to development to
test it. Ensure that search is working and that data is correct on the development
system.
- Test everything on production, even if it replicated properly to development.
- Do not run multiple replication processes at the same time.
- If you modify a system object type, replicate global system object types before
replicating any objects of that type.
- Identify and follow the dependencies between different data replication groups. For
example, if you have campaigns that use source codes and coupons, replicate source codes
and coupons before, or together with, campaigns. Replicating campaigns first in that
situation can result in data corruption.
- When replicating to a production instance, first transfer the data or code to verify
the transfer process. Examine the logs before publishing the data or activating the
code.
- Always rebuild your search indexes and make sure that the process is complete before
replicating them. When replicating search indexes, disable incremental and scheduled
indexing, and stop other jobs. Replication fails when an index is being rebuilt or
modified.
- Replication processes are resource-intensive and can affect site performance. Run
replication during low traffic times, such as late night or early morning hours.
Schedule replication to production when storefront activity is at a minimum. Try to
avoid major data replications during Salesforce B2C Commerce's standard maintenance
windows.
- Limit user privileges to perform data replication, and split the responsibility between
multiple user roles. For example:
- A product manager defines replication processes, but does not have permission to
schedule or execute them.
- A replication manager manages and executes replication processes.
- Avoid modifying or moving static content files whenever possible. These changes take a
long time to replicate.
- Copy an existing replication process when possible, instead of creating a process from
scratch.
- When the replication process automatically clears the page cache, don't clear it again
manually.
- Do not edit cartridge paths or code directly in a production instance. Doing so can
result in unexpected behavior, such as multiple versions of pages. Always replicate code
and preferences from staging to production.
- For performance reasons, we recommend you store no more than 100,000 objects (regular
files and directories) in a directory.