nsqd: messages defer time lost after restart

[chyroc]

1. start nsq
2. pub 100000 defer message(these message well store in file by go-diskqueue)
3. stop nsq(lost defer-time info? because file only store message itself, not contain defer-time)
4. start nsq again (all messages will restored from file and consume these message immediately)

also store defer-time to file?

[ploxiln]

Right - this is a known limitation of nsqd. It was not designed to be a robust scheduler. It can still be useful to build a system around - e.g. maybe the message has the time it should be processed, and the consumer can re-queue with appropriate delay if it gets it early. It works better for shorter defer times, like just a few minutes.

This is something that could theoretically be fixed, so I'll leave this open :)

[mreiferson]

Re-labeling this. Although it's technically an "issue", it could easily be treated as "working as intended" and not a "bug".

[mreiferson]

see #1170, which would fix this

[qaqhy]

fix: #1454 Fixed the issue of message delay time loss caused by nsqd restart