Hi Tiansheng, Thanks for the reporting; In 1.5.3 there is some code refactoring, but the cluster deployment method didn't change: only one node be "all" and others be "query". For the error message "scheduler has not been started" in query node, it is a defect (will create a JIRA for that), but that doesn't impact the function.
For the second case: making all nodes as "all", there should be only 1 running the job engine, others saying "fail to acquire lock, scheduler has not been started", is that true in your case? Because the zookeeper is used as a lock to avoid two job engines be started. 2016-08-02 20:43 GMT+08:00 张天生 <[email protected]>: > Btw, i deployed the newest version kylin 1.5.3 version. > > 张天生 <[email protected]>于2016年8月2日周二 下午8:41写道： > >> I deployed kylin in cluster mode. There are 1 "all" server and 2 "query" >> servers. When i started query server, the following error log appeared: >> >> 2016-08-02 20:26:55,855 INFO [Thread-11] threadpool.DefaultScheduler:178 >> : server mode: query, no need to run job scheduler >> 2016-08-02 20:26:55,856 ERROR [Thread-11] controller.JobController:91 : >> scheduler has not been started >> 2016-08-02 20:26:56,856 ERROR [Thread-11] controller.JobController:91 : >> scheduler has not been started >> 2016-08-02 20:26:57,856 ERROR [Thread-11] controller.JobController:91 : >> scheduler has not been started >> 2016-08-02 20:26:58,856 ERROR [Thread-11] controller.JobController:91 : >> scheduler has not been started >> 2016-08-02 20:26:59,857 ERROR [Thread-11] controller.JobController:91 : >> scheduler has not been started >> >> However, when i deployed 3 server in "all" kylin.server.mode, it worked >> fine. But offical document said it only one server can run the job >> engine(“all” mode or “job” mode), the others must all be “query” mode.I >> don't know why, can someone tell me the resson? >> >> -- Best regards, Shaofeng Shi
