ConnectionPool::PoolShuttingDownError

[sahin]

If there is a network error or connection error, and if there is a change replication set primary, we started to have this error. It happens like every 10-12 hours and create thousands of errors. We have to app restart to fix it.

I am using latest moped 2.0.5.

[nofxx]

They are all related, this one, #346, #348 ...
Default pool size (5) start the failures in a few minutes, experimenting with 70 gives some days.

[nofxx]

@sahin do you have many #find_and_update on your code? It's the only change, you might try mongoid 5, and get rid of moped.
Thinking about doing it now, we're in a fork in mongoid's life, suffering in the transition heh.
Also, which mongodb version?

[sahin]

@nofxx , we are in mongo 4.0.2

our pool is 500. We have instant 100 connections from web, api, player. and workers.

We never user find and update. We have lots of find, user related data writes , and many writes in workers, there are find, update or create.

I would loce to get rid of moped. Is mongoid 5 ready to use in production?

I would Is mongoid 5

[sahin]

We use mongo 2.6.10 . We will move to 3.0.3 soon.

I would love to use mongoid 5

[nofxx]

Hello sahin. It's not production ready, check the repo, but can't be worst than this pool nightmare.
Or it can. If you code doens't have any find_and_update, just change mongoid on Gemfile to use git:

Here we are with mongo 3.0.3, moped 2.0.5 patched with pool_size increase.
But I still got some pool issues some hours ago, hoping for a day w/o errors...
If it happens at work hours today going to try mongoid 5 tomorrow.. =/

[sahin]

"hoping for a day w/o errors..." this connection pool error makes me wake up 2-3 times at night....

@arthurnn @durran do u guys suggest to move mongoid 5 master

[nofxx]

Without any rep machine failing it's more than a day w/o errors now. Not sure if one goes down, will try sometime in the early morning today.

Only issue now is a sidekiq worker that won't stop eating RAM. Not sure it's related to the moped pool, experimenting now.

[agrass]

Anyone here who already tried mongoid5? @sahin? We are with mongoid 4.0.2 and moped 2.0.6 and still failing

[marlonoso]

+1 same here

• mongoid (4.0.2)
• moped (2.0.6)
• pool_size: 100

[jlhonora]

We just migrated to rails 4.1.12 and the issue seems to be gone (+2 weeks, never saw the bug since). No idea how it could be related.

[nofxx]

Now that you mentioned rails @jlhonora , might be what fixed here too, but we're in 4.2.3.
What we still get is sidekiq not releasing RAM, it happened together with the pool bug, but now the pool bug is gonne and the ram eating persists.

[brand-it]

I am going to try going from 4.1.10 to 4.1.12 and see if that changes anything. I will post if I see the same results as @jlhonora.

[brauliobo]

same here, no fix?!?

[brand-it]

I have not seen the bug after going to gem 'mongoid', '5.1.1'