Mongos and broken pipe exceptions after repairDatabase

[Justin Miller]

I'm repairing my replica sets according to http://www.mongodb.org/display/DOCS/Durability+and+Repair#DurabilityandRepair-RepairCommand. I first step down the masters, restart, repair, etc., then I do the secondaries. 

In my script, to speed up this process, I actually do the repairs in parallel. First I go through and build a list of who my masters are and who my secondaries are. I start with the masters and execute the procedure in parallel. Then I do it for the secondaries after the (previous) masters have come back up.

I notice though, that when all is said and done, my application, specifically the mongos's seem to have problems reconnecting to the shards and continue to throw application errors. I see this in the mongos logs:

"warning could not clear last error from shard ... caused by ... socket exception"

"Socket say send()  errno:32 Broken pipe ..."

If I restart my mongos's all is well, but this isn't ideal.

We're using Mongodb 2.0.4. 

Thoughts?

Thanks,

Justin

[Kristina Chodorow]

At the moment, the only way to refresh them all is, as you discovered,
restarting the mongos processes. There's a server issue for this you
can follow, though: https://jira.mongodb.org/browse/SERVER-4706.


On Apr 24, 4:19 pm, Justin Miller <justin.adam.mil...@gmail.com>
wrote:
> I'm repairing my replica sets according tohttp://www.mongodb.org/display/DOCS/Durability+and+Repair#Durabilitya....