3 ms·
After the horrendous time I've had with "modern" MongoDB losing my data repeatedly through 2018, I will never touch it again. Postgres handles loss of power, fu
by StudentStuff 8y ago
After the horrendous time I've had with "modern" MongoDB losing my data repeatedly through 2018, I will never touch it again. Postgres handles loss of power, full disks, kill -9 and numerous other edge cases just fine, while MongoDB proceeds to corrupt its data and not fail gracefully.
Even the redheaded stepchild of databases that is MSSQL goes into read only mode when you have a full disk, there is no excuse for Mongo corrupting itself when a disk is full.
- gnulinux 8y agoNot a formal criticism because we didn't investigate too much into this, but just last week our dev db instance run out of disk and postgres did some really suspicious things. We confirmed that on multiple occasions queries returned with success (because we have logs of success) but the db writes weren't persisted. So the software assumed data was written and continued on happy path, leaving the db in a bad state. Luckily this was only a dev machine and trivial to recover from, and granted out of disk is a fatal problem, but we were still upset.
- StudentStuff 8y agoWe've seen Postgres continue to operate like this until you run out of ram, not ideal IMO, but better than corrupting the existing database. Ideally Postgres would lift that whole "on full disk go read only" feature from MSSQL.
- anarazel 8y agoDid you turn fsync off?