Balancer stuck due to config server crash

20 views

Skip to first unread message

Darshan Shah

unread,

Apr 22, 2016, 11:53:49 AM4/22/16

to mongodb-user

I am running a MongoDb 3.2.5 tag-aware sharded cluster with WiredTiger having 3 config servers (which are not in a replicaset) and each node being a 2 node replicaset.

In the sharded cluster, the config servers were restarted while the balancer was balancing a tag-aware collection.

After the config servers were restarted, I see that the balancer is stuck on balancing that same collection for over few hours.

I manually removed the balancer lock and then it proceeded with balancing other collections and still keeps showing that particular collection as also being migrated.

Is this behavior normal?

How do we cleanly recover in such a scenario?

Thanks!

Kevin Adistambha

unread,

Apr 29, 2016, 1:41:05 AM4/29/16

to mongodb-user

Hi Darshan

I manually removed the balancer lock and then it proceeded with balancing other collections and still keeps showing that particular collection as also being migrated.

Have this issue been resolved? Or is this is still happening?

If you think this is still an issue, could you tell us:

the output of sh.status()
what you mean by “showing that particular collection as also being migrated.”

Since config servers are absolutely vital to the operation of a sharded cluster, I would recommend you to upgrade your config servers to a replica set to improve performance and consistency on the config servers.