Need citation. ;) May you already know, you can easily implement heartbeat with master (heartbeat checker) process - just leave process' information and timestamp. Your master process checks it periodically and kill process if needed. (If you scales out to multi nodes, each node should have master process to kill local process.)
Zookeeper is for H/A, so if you start with just one node, it would be fine to start with easiest solution (non H/A), and apply H/A if it scales to multi nodes. (If that situation, Zookeeper would be fine. :) ) About storage, it would be depending on your heartbeat rate and all processes count, but actually most of storage is fine to use. In-memory storage could be also fine because heartbeat information would be small and we only need to storage last heartbeat from each process. Please correct me if I'm wrong. Thanks! Regards. Jungtaek Lim (HeartSaVioR) 2014-11-19 18:36 GMT+09:00 M.Tarkeshwar Rao <[email protected]>: > Thanks for reply. > --- 1. What's the reason your team decides to find alternatives other than > Zookeeper? > > Actually our customers are not ready to go for more than two nodes and zoo > keeper require more than 3 nodes for cluster setup. > > ---2. What / how do your new project synchronizes? > all the processes commnicating on using tcp sockets. but now we going for > more than one node. For this we need very basic feature of zookeeper so > that we can implement heartbeat like storm is doing when any worker or > supervisor killed. > > ---3. What's CFS, and what if your single machine running synchronization > service has failure? Your distributed components can also having issues so > it's important. > > CFS means Cluster File System. It will be the central Disk shared between > two or more nodes. For minimal change in the present code base we are going > for this design. > > Regards > Tarkeshwar > > > On Wed, Nov 19, 2014 at 12:21 PM, 임정택 <[email protected]> wrote: > >> Hello. >> >> Just a curious, some question regarding on this. >> >> 1. What's the reason your team decides to find alternatives other than >> Zookeeper? >> 2. What / how do your new project synchronizes? >> 3. What's CFS, and what if your single machine running synchronization >> service has failure? Your distributed components can also having issues so >> it's important. >> >> 2014-11-19 15:13 GMT+09:00 M.Tarkeshwar Rao <[email protected]>: >> >> >>> Hi all, >>> >>> We are using storm in one of our project. It is suitable for us. >>> >>> But one of my project need distributed processing but we are not using >>> storm in that. >>> >>> But we take the basic design from the storm. Thanks for good distributed >>> design that Storm has. >>> >>> For synchronization storm using Zookeeper. >>> >>> This new project is written in c++. We want to use Zookeeper like >>> service for synchronization which can be deployed on single machine on >>> CFS. >>> >>> Can you please provide your inputs? >>> >>> Regards >>> Tarkeshwar >>> >>> >> >> >> -- >> Name : 임 정택 >> Blog : http://www.heartsavior.net / http://dev.heartsavior.net >> Twitter : http://twitter.com/heartsavior >> LinkedIn : http://www.linkedin.com/in/heartsavior >> > > -- Name : 임 정택 Blog : http://www.heartsavior.net / http://dev.heartsavior.net Twitter : http://twitter.com/heartsavior LinkedIn : http://www.linkedin.com/in/heartsavior
