Jonathan Hyman
Jonathan Hyman
We experience this issue as well. Would love to see it merged.
@telent, have you been running this in production? I tested this out on production and had some really strange errors (moped cursor returning empty documents) that were sporadic and hard...
Though to be clear, I had done it by copying the code into a fork of Moped 1.5. I didn't try with Moped 2 yet.
I tried testing this again by issuing a scatter-gather query on a massive collection with hundreds of millions of documents. This wreaked havoc on the database, as it appears that...
@durran have you thought about this problem at all in the new ruby driver? As @telent points out, `SO_RCVTIMEO` simply does not work in MRI anymore. We experience this issue...
I may be wrong, but I don't think that that's going to help. Going back to Moped and this issue for a second, the Moped code already has a `check_if_alive!`...
Is your replica set or network connection failing at some point? See discussion at https://github.com/mongoid/moped/pull/331
Kiqstand doesn't need to (and in my opinion, should not) disconnect anything. Sidekiq's workers are in fiber-local variables, they will disconnect automatically at the end of the job.
See https://github.com/mongoid/kiqstand/pull/10, and for a fuller explanation, see https://github.com/mperham/sidekiq/issues/1526
@arthurnn can you take a look? We just upgraded to Moped 2 in production last night and have been wrecked by this bug so far.