Ubuntu 18.04 LTS server, disk is ssd. Application stack is: mongodb, node, nignx and some python. Server is up 1 month and already had 3 OS disk fails where Ubuntu would go in read only mode. After restart in boot mode it would show invalid checksum recovering blocks. It would be resolved with fsck command.
From amplify (nginx logging tool) I see I/O latency spikes with no extra cpu, or traffic present. I did run smrtctl 4 times and it says disk has no errors. What could be the coulprint for spikes in disk latency (this is than sometimes shown in slow website response)? In attachment are logs for last week, and last 4h:
