diff options
| author | Simon MacMullen <simon@rabbitmq.com> | 2014-10-20 20:57:13 +0100 |
|---|---|---|
| committer | Simon MacMullen <simon@rabbitmq.com> | 2014-10-20 20:57:13 +0100 |
| commit | 50c3620242aba4aafd19ce16edbdc1faefb745a5 (patch) | |
| tree | 46a115a42696991397e94453f0cd8e8cec971fc5 /src/gm.erl | |
| parent | 183a1ab479d409db0e9ffdcbd6e9ee7422b0b119 (diff) | |
| download | rabbitmq-server-git-50c3620242aba4aafd19ce16edbdc1faefb745a5.tar.gz | |
Sleep briefly, and explain why.
Diffstat (limited to 'src/gm.erl')
| -rw-r--r-- | src/gm.erl | 8 |
1 files changed, 8 insertions, 0 deletions
diff --git a/src/gm.erl b/src/gm.erl index 6c09e66417..636e63e4b7 100644 --- a/src/gm.erl +++ b/src/gm.erl @@ -720,6 +720,14 @@ handle_info({'DOWN', MRef, process, _Pid, Reason}, {_, {shutdown, ring_shutdown}} -> noreply(State); _ -> + %% In the event of a partial partition we could see another member + %% go down and then remove them from Mnesia. While they can + %% recover from this they'd have to restart the queue - not + %% ideal. So let's sleep here briefly just in case this was caused + %% by a partial partition; in which case by the time we record the + %% member death in Mnesia we will probably be in a full + %% partition and will not be assassinating another member. + timer:sleep(100), View1 = group_to_view(record_dead_member_in_group( Member, GroupName, TxnFun)), handle_callback_result( |