0

I've had problems with a MongoDB replica set of my sharded cluster. The current status of the nodes is:

  • node1: primary (but apparently vetoed by node3 that needs to sync)
  • node2: rollback (it was the primary before)
  • node3: can't sync
  • node4: delayed slave currently recovering
  • node5: arbiter

The documentation seems to say that I have to wait the rollback to be finished and to restore or not the dump on the new primary. But the node2 has already been doing the rollback for half an hour and I'm wondering if it's stuck?

What are the steps I should follow? Can you give me some light on the way to restore the rollback-dump on the new primary?

Thanks a lot,

Greg.

Icu
  • 1,405
  • 2
  • 15
  • 25

1 Answers1

0

I let the node2 doing its rollback for 2 hours then stopped it. I couldn't find any rollback directory in dbpath ...

As the processes feeding the DBs have a internal buffer mechanism, they can re-transmit recent data. So in the case I loose data, I can retrieve it again.

Therefore, I choosed to stop all the nodes of the replica set. Then I started the old primary as a standalone server, I dropped the local database and deployed a new replica set from there.

Greg.

Icu
  • 1,405
  • 2
  • 15
  • 25