Hacker News new | past | comments | ask | show | jobs | submit login

While I want to agree, until this year there were still deadlock issues that could take down whole clusters [0][1].

[0] https://jira.mongodb.org/browse/SERVER-57476

[1] https://jira.mongodb.org/browse/SERVER-56779




A JIRA user is needed to see them I guess. Do you mind to tell what are the issues ?


I could open the issues in icognito mode. 1: Operation may block on prepare conflict while holding oplog slot, stalling replication indefinitely

2: Do not use the collection distributed lock for chunk merges


Yep. #2 is kind of a hush-hush way of them fixing a bug that can take a lock on a collection during a chunk merge and fail to release it, resulting in writes to primaries to stop until they are kill -9'd.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: