If periodic processor killed,the cron info will be lost？

[mad3310]

I want to know that how to slove the cluster problem with periodic? cron info need to store in etcd? DB?

[Lupino]

Just store data on leveldb or redis

periodic -d --dbpath=/path/to/leveldb --driver leveldb
# or
periodic -d --reids=your_redis_host --driver redis

more help periodic -h

[mad3310]

If the two periodic processor use the cron info, how to slove only one processor to dispatch this work?

[Lupino]

If Two periodic servers use the same job info, there will do same job twice.
from the client cluster make sure that different servers doing the different tasks.

[mad3310]

According to your idea, one periodic server has different task. If the server shut down, these tasks will not be pushed to associated worker?

[Lupino]

In my case I use one periodic server, it auto restart by supervisord and store data on leveldb,

maybe use the master slave model, the two periodic server connect one redis server, all the client connect the master. if master crashed all client switch to slave, and the slave become a master

[mad3310]

In my opinion, the best way is horizontal scale. If a lot of tasks, one periodic server may be diffcult to dispatch these work. any sounds?

[Lupino]

You'd do it youself. one solution is: add a proxy server before periodic server. in the proxy server, you choose the message send to which periodic server.