2015-08-20 01:11:39 |
Liusheng |
description |
In devstack environment with mysql storage backend, following steps to reproduce this issue:
1. drop all db tables
2. do ceilometer-dbsync.
3. config the coordination backend_url as "backend_url=file:///opt/stack/partitioncoordinator/lock"
4. restart ceilometer services (include ceilometer-agent-notification and ceilometer-polling)
the "ceilometer meter-list" command will return a empty list.
when I restart agent-notification service with an separated config file without partition backend_url, it can works ok. |
In devstack environment with mysql storage backend, following steps to reproduce this issue:
1. drop all db tables
2. do ceilometer-dbsync.
3. config the coordination backend_url as "backend_url=file:///opt/stack/partitioncoordinator/lock"
4. restart ceilometer services (include ceilometer-agent-notification and ceilometer-polling)
the "ceilometer meter-list" command will return a empty list.
when I restart agent-notification service with an separated config file without partition backend_url, it can works ok.
---------------------------------------------------------------
The main reason of above issue is because I have configured the backend_url of partition_coordinator as file driver. after restarting the ceilometer-polling service several times, the polling-agent cannot poll resources, because with file driver, the member of partition group will still exist even the member leaved after restarting service. the member of tooz partition with filter driver is show as a file, the file won't be removed after group member leaved. |
|