support non-clean shutdown

[tsahee]

uses geth shutdown markers (requires stopping blockchain before backend)

logs time for block creation fo blockchain will write trie cache

Pulls in OffchainLabs/go-ethereum#155

[joshuacolvin0]

LGTM

[codecov]

Codecov Report

Merging #1039 (f1a7d09) into master (f68775b) will decrease coverage by 0.27%.
The diff coverage is 42.42%.

@@            Coverage Diff             @@
##           master    #1039      +/-   ##
==========================================
- Coverage   48.03%   47.75%   -0.28%     
==========================================
  Files         218      218              
  Lines       25180    25207      +27     
  Branches      523      523              
==========================================
- Hits        12094    12037      -57     
- Misses      11477    11538      +61     
- Partials     1609     1632      +23     

[PlasmaPower]

LGTM

Dismissing approval as Harry has some suggestions for the geth side

[PlasmaPower]

Let's update this to an hour and then LGTM :)

[PlasmaPower]

I've pushed up a commit changing this to an hour

[joshuacolvin0]

LGTM

[hdiass]

This only works if node works for more than 1 hours.. not an amazing solution as it can be in sync for less than a hour thus if restarted in less than 1 hour state gets bricked

[PlasmaPower]

@hdiass this is configurable with --node.caching.trie-time-limit=30m for instance to make it 30 minutes, but be aware that lowering the time to sync after an unclean shutdown increases the db size. We follow upstream geth behavior here.