After upgrading all OSDs to bluestore, the issue resolved.
One of OSD had filestore format while others had bluestore. This may be related to the issue.
No more "division by zero" message and data is actually sent by zabbix_sender, verified by using tcpdump.
2017-09-22 23:04:05.957635 7fcb03ae9700 20 mgr[zabbix] Waking up for new iteration
2017-09-22 23:04:05.958547 7fcb03ae9700 20 mgr[zabbix] Sending data to Zabbix server 10.18.10.18
2017-09-22 23:04:05.958625 7fcb03ae9700 20 mgr[zabbix] {'rd_bytes': 11398143492096L, 'total_bytes': 4798027112448L, 'overall_status_int': 1, 'osd_latency_apply_min': 0L, 'osd_latency_apply_max': 5L, 'num_osd_up': 6, 'osd_max_fill': 13.095749553280234, 'osd_backfillfull_ratio': 0.8999999761581421, 'osd_latency_commit_max': 5L, 'wr_bytes': 3337205972992L, 'osd_latency_commit_min': 0L, 'num_mon': 3, 'osd_min_fill': 2.037110421878282, 'total_used_bytes': 359502389248L, 'wr_ops': 305991965L, 'osd_nearfull_ratio': 0.8500000238418579, 'osd_latency_apply_avg': 1.0, 'overall_status': u'HEALTH_WARN', 'num_pg': 64L, 'osd_latency_commit_avg': 1.0, 'osd_avg_fill': 7.492748589967889, 'num_osd': 6, 'osd_full_ratio': 0.949999988079071, 'total_objects': 28475L, 'num_pools': 1, 'num_osd_in': 6, 'num_pg_temp': 24, 'rd_ops': 181606152L, 'total_avail_bytes': 4438524723200L}
2017-09-22 23:04:05.965294 7fcb03ae9700 0 mgr[zabbix] Exception when sending: /usr/bin/zabbix_sender exited non-zero: zabbix_sender [9297]: DEBUG: answer [{"response":"success","info":"processed: 0; failed: 29; total: 29; seconds spent: 0.000180"}]
2017-09-22 23:04:05.965396 7fcb03ae9700 20 mgr[zabbix] Sleeping for 60 seconds