todo.txt

- log to mongodb!
- finish election
- send leader address to client
- do two phase commit
- write log file
- sync total data
- sync partial data
- encode values in a better way
- use logrotation and compression!


catch nonode errors in getchildren


tests:
	assure that biggest logfile wins!
	run 100 client requests,
	make sure everybody has same log file size
	check kazzoo connection, deal witch suspend and lost - how?
	set status to down if commit failed!

make sure client 1 does not accept prep_requests after 1000 requests
bring client back after number of requests

deal with lockfiles!
./db0/LOCK
- does client need to contain a database?

better communication with client, return tuple of what to do!
order of read and write requests

close filehandler!!!
if i am first child i have to be primary!

testcase:
	primary alone, deny, add a backup, accept

check master failure

clear database if child!

do selftest logfile == consecutive logids, key, value checksum


check different use cases:
	startup:
		if !is primary:
			get op log and replay database
			set ephemeral node or notify primary zerorpc



	Replica errors:
	ready fails
		see if child is alive - retry

	commit fails
		identify child:
			if child is still alive, restart child remotely
			else
			wait for child to restart and sync operations log

	Primary error:

	everybody child watches on ready
		new node comes up
		gets operations log from master
		updates database
		sets ephemeral node