Replicating to DR from PRD

andrewl657 over 7 years ago

We have a distributed environment set up for our Production instance. F5 load balanced to 2 web servers, 2 app servers (active-passive), & clustered db. Our DR, on the other hand, is F5 to 1 web server, 1 app server, & 1 db server. Looking through the documentation on backup/replication, it mentions that it is necessary to have the exact same setup/configurations between DR and PRD.

We are not able to do that because our domain name and url for the DR will be different so we are not able to replicate the whole <APPIAN_HOME> directory. We have all engines & app data on a SAN that we plan to backup and replicate to our DR including RDBMS. Does anyone see any problems with this since the application setup will be different? Also I will be running the checkpointing script before every backup. Should we only be replicating the top 2 highest kdb files over to DR along with all app data and RDBMS?

Th...

Discussion posts and replies are publicly visible

0 andrewl657 over 7 years ago

...anks.
Cancel
Vote Up 0 Vote Down

Sign in to reply

Verify Answer

Cancel
0 Mike Cichy
Appian Employee
over 7 years ago

The reason for the recommendation is so that the DR site can handle PROD load but from your description it sounds like it would be.
You do have to have the same software configuration on the DR machine (e.g. the number of process execution engines).
I do not see a problem with replication, and 2 highest kdbs (current + 1 closed one) is sufficient. Are you replicating only the backed up kdbs? How frequently are you doing that? Is it going to satisfy your RPO?
Cancel
Vote Up 0 Vote Down

Sign in to reply

Verify Answer

Cancel
0 andrewl657 over 7 years ago

Hey Mike. Thanks. We will have the same number of engines on both. We will be replicating only backed up kdbs and this will be every 4 hours. We are trying to avoid more than 4 hours of data loss but I am concerned that this frequency could cause performance issues in the future if the kdbs grow in size. What are your thoughts?
Cancel
Vote Up 0 Vote Down

Sign in to reply

Verify Answer

Cancel
0 Mike Cichy
Appian Employee
over 7 years ago

Checkpointing does prevent writes to kdbs so I would pick times of low usage to do the checkpoint. I am not sure what your data load projections are, but the size of the kdbs can become a concern but likely can be deferred for a while.
The proposed approach seems sound. You do have an option of replicating "live" kdb files but that will increase your RTO, since the transaction log would have to be replayed on DR site startup and I would not get into it unless your checkpointing times are really long.
Cancel
Vote Up 0 Vote Down

Sign in to reply

Verify Answer

Cancel