Skip to content
This repository was archived by the owner on Oct 22, 2021. It is now read-only.
This repository was archived by the owner on Oct 22, 2021. It is now read-only.

Cluster dies on mem3_rep_manager #106

@opie4624

Description

@opie4624

Yesterday a single node was unable to start due to errors in '''mem3_rep_manager''' after a few hours, all 6 nodes are unable to start.

Each node's run, right now, looks like this:

[Tue, 22 May 2012 19:04:07 GMT] [info] [<0.87.0>] [--------] Apache CouchDB has started on http://undefined:5986/
[Tue, 22 May 2012 19:04:08 GMT] [error] [emulator] [--------] Error in process <0.171.0> on node 'bigcouch@couchdb1' with exit value: {{badmatch,nil},[{fabric_view,remove_down_shards,2},{rexi_utils,process_mailbox,6},{fabric_view_changes,receive_results,5},{fabric_view_changes,send_changes,6},{fabric_view_changes,go,5}]}


[Tue, 22 May 2012 19:04:08 GMT] [error] [<0.164.0>] [--------] ** Generic server mem3_rep_manager terminating 
** Last message in was {'EXIT',<0.171.0>,
                               {{badmatch,nil},
                                [{fabric_view,remove_down_shards,2},
                                 {rexi_utils,process_mailbox,6},
                                 {fabric_view_changes,receive_results,5},
                                 {fabric_view_changes,send_changes,6},
                                 {fabric_view_changes,go,5}]}}
** When Server state == {state,<0.165.0>,10,nil,[<0.171.0>]}
** Reason for termination == 
** {unexpected_msg,{'EXIT',<0.171.0>,
                           {{badmatch,nil},
                            [{fabric_view,remove_down_shards,2},
                             {rexi_utils,process_mailbox,6},
                             {fabric_view_changes,receive_results,5},
                             {fabric_view_changes,send_changes,6},
                             {fabric_view_changes,go,5}]}}}

The last http request was retrieving a view, then a stack trace just like above happened and the node went down. Since the load balancer would kick each request to a working node, that eventually downed the entire cluster.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions