Failed elasticsearch upgrade 2.3.4 -> 5.5.0 , all shards is unassigned

[EnglishVillage]

es2.3.4 -> es5.5.0 upgrade,shutdown old cluster and reboot new cluster, new cluster is running, but all shards is unassigned.
env:

• es2.3.4:delete-by-query, head, jcseg, kopf, license, marvel-agent, sql
• es5.5.0:sql, head, cerebro
• java version "1.8.0_111"
• centos7.2
  
  

[Igor_Motov]

A log file from the master node or output of allocation explanation api would really help us in identifying the issue.