Nodebb: analytics duplicate key with clustered instance

[BenLubar]

24/3 09:50 [100] - error: [analytics] Encountered error while writing analytics to data store: E11000 duplicate key error index: 0.objects.$_key_1_value_-1 dup key: { : "analytics:pageviews:month", : "1456790400000" }
...
24/3 10:10 [100] - error: [analytics] Encountered error while writing analytics to data store: E11000 duplicate key error index: 0.objects.$_key_1_value_-1 dup key: { : "analytics:pageviews:month", : "1456790400000" }
...
24/3 12:10 [100] - error: [analytics] Encountered error while writing analytics to data store: E11000 duplicate key error index: 0.objects.$_key_1_value_-1 dup key: { : "analytics:pageviews:month", : "1456790400000" }

I'm pretty sure this is because the 4 instances all tried to write the analytics data as if they were the only instance.

[barisusakli]

Yeah each instance counts its own pageviews, and every 10mins they all try to increment the same key. Pretty sure this is caused by a mongodb bug Until that is fixed we can add some code so the updates don't run at the same time.

[barisusakli]

Let me know if it happens again, the mongodb docs suggest retrying failed upserts on unique indices.
https://docs.mongodb.org/manual/reference/command/findAndModify/#upsert-and-unique-index

To prevent the creation of multiple duplicate documents, create a unique index on the name field. With the unique index in place, then the multiple findAndModify commands will exhibit one of the following behaviors:

Exactly one findAndModify successfully inserts a new document.
Zero or more findAndModify commands update the newly inserted document.
Zero or more findAndModify commands fail when they attempt to insert a duplicate. If the command fails due to a unique index constraint violation, you can retry the command. Absent a delete of the document, the retry should not fail.

I guess that's why you only see 3 errors with 4 instances, 1 of them succeeded the other 3 failed so retrying the 3 will succeed since the document is now in the database.