travelping / dike Goto Github PK

Not all processes from dike are proper started within supervisor tree. This should be fixed, in a way, that all processes are in supervisor tree, and if some error is happen, the process should be restartable.

when a node is interrupted in its recovery process, another node will not release a lock.
by this mechanism, the whole cluster can be locked

I have to switch parse_transform and lager_transform in config files before the installation following the readme is possible.
Is that intended?

Easy reproducable with dike_cache :

Start 5 nodes like:

application:load(dike),
application:set_env(dike, masters, Nodes),
application:set_env(dike, paxos_timeout, 1000),
application:ensure_all_started(dike),
dike_cache_api:new(dike_cache, test_cache, 256).

where Nodes are 5 nodes on your local machine.

After that, restart 1 node, the restart of one node(with the same sequence) takes more than 12 minutes(do not try to wait longer), with no load, with no messages.