This reverts commit 1189a4ea.

Actually this is wrong, the command can be called without args at all.

connection to master/slave/sentinel decames disconnected just after the last PONG and before the next PING.

1 microsecond = 1000 nanoseconds
1e3 = 1000
10e3 = 10000

This fixes a bug introduced by d827dbfd, and makes the code consistent
with the logic of always allowing, while the cluster is down, commands
that don't target any key.

As a side effect the code is also simpler now.

Fix issue in case the redirect address is in ipv6 format. Parse from behind to extract last part of the response which represents actual port.

Thanks to @tushar2708 for the PR. I applied a slightly different fix.
Thanks to @cespare for reporting.

Close #3024
Close #3020

Fixes #2515.

I've renamed maxmemoryToString to evictPolicyToString since that is
more accurate (and easier to mentally connect with the correct data), as
well as updated the function to user server.maxmemory_policy rather than
server.maxmemory. Now with a default config it is actually returning
the correct policy rather than volatile-lru.

Thanks to @oranagra for the idea of allowing CONFIG GET during loading.

Thanks to @oranagra for the hint about misplaced 'r' flags.

Thanks to @tidwall for reporting.
Close #3194.

The test works but is very slow so far, since it involves resharding
1/5 of all the cluster slots from master 0 to the other 4 masters and
back into the original master.

This fixes issue #3043.

Before this fix, after a complete resharding of a master slots
to other nodes, the master remains empty and the slaves migrate away
to other masters with non-zero nodes. However the old master now empty,
is no longer considered a target for migration, because the system has
no way to tell it had slaves in the past.

This fix leaves the algorithm used in the past untouched, but adds a
new rule. When a new or old master which is empty and without slaves,
are assigend with their first slot, if other masters in the cluster have
slaves, they are automatically considered to be targets for replicas
migration.

As a side effect, cat commands.txt | redis-cli now is able to handle
lines more than 4096 bytes.

This fix was written by Anthony LaTorre.
The old code mis-calculated the amount of time to wait till next event.

This fix was suggested by Anthony LaTorre, that provided also a good
test case that was used to verify the fix.

The problem with the old implementation is that, the time returned by
a timer event (that is the time after it want to run again) is added
to the event *start time*. So if the event takes, in order to run, more
than the time it says it want to be scheduled again for running, an
infinite loop is triggered.