Consider this scenario:
- Two identical networks (each device and link is exactly duplicated). Geographically diverse.
- One stand-alone DMA in each network (different DMA IDs.) No communication at all between them.
- One network & DMA is active/primary; the other is stand-by/secondary
- All secondary devices mirror the IP address of their primary twin
- A backup of the primary DMA is taken every night
- If the primary network goes down, the backup is restored on the secondary DMA and the secondary network becomes active.
- The secondary DMA is clean before restoring the backup
Are there any other considerations to keep in mind for this to work and have the secondary network take over without a hitch?
Marieke Goethals [SLC] [DevOps Catalyst] Edited question 6th August 2020
At first sight this seems a perfect cold redundant setup. A few things that come into my mind that you need to take into account:
- IP of the secondary DMA. If your products are sending traps towards DataMiner. You need to be make sure that after a switch to the secondary DMA, they are still sending it to the correct DMA. If the IP of the main and secondary DMA will be the same, then there's no problem.
- The license. A DMA license is created per DMA ID and MAC address. The secondary DMA will have another MAC and DMA ID. So make sure you have a .lic file available at the time you want to switch to the secondary.
- Switch time. You're saying 'take over without a hitch'. Take into account that restoring a backup might take a couple of minutes. It's like running an upgrade package on a DMA. Compared to a hot redundant system, this will be slower.
- Security. If you're using domain accounts to connect do DMA. Make sure that in the secondary network, the DMA has access to the same domain. Otherwise you might end up in a situation that the users that are restored on the secondary DMA have no access to the system.
- Time zone. Make sure that the time of the main and backup DMA are synced/are on the same time zone to avoid issues like alarms/trending that would suddenly appear in the future.
- DataMiner version. At all times the main and (empty) backup DMA need to be on the same DataMiner version. When upgrading the main, don't forget to also upgrade the backup in the other network. If the network setup permits , you might cluster the backup agent and the main agent 'as external DMAs'.
- Double polling of products. Make sure that if you spin up your secondary DMA, you're main DMA is stopped to avoid double polling of your products. Double polling might cause problems for certains equipment.
- Storing the backup pacakage. Might sound obvious, but don't store your backup package on the same machine where the main DMA runs on. If it goes down, you're backup package to restore is gone as well.
Jorge Sienra [SLC] [DevOps Enabler] Posted new comment 30th July 2020
Thank you Jochen,
Since there are two identical networks, each one talking to just one DMA, I guess there is no issue with each DMA having its own IP address? all devices in each network will send the traps to the IP of their own DM.
And we wouldn’t have the risk of double-polling either.