diff options
| author | Simon MacMullen <simon@rabbitmq.com> | 2015-02-04 12:06:44 +0000 |
|---|---|---|
| committer | Simon MacMullen <simon@rabbitmq.com> | 2015-02-04 12:06:44 +0000 |
| commit | 0688423a2d506d599608d94ba7231864592d1427 (patch) | |
| tree | 00bd96ad4d213a0dd1a5a7585548b445ddd7c422 | |
| parent | 2e73febaebcc4f64a2306028834bd0f238346b06 (diff) | |
| download | rabbitmq-server-git-0688423a2d506d599608d94ba7231864592d1427.tar.gz | |
A bit more explanation.
| -rw-r--r-- | src/worker_pool.erl | 5 |
1 files changed, 5 insertions, 0 deletions
diff --git a/src/worker_pool.erl b/src/worker_pool.erl index 06a670ff4c..eb8c1dc6b8 100644 --- a/src/worker_pool.erl +++ b/src/worker_pool.erl @@ -28,6 +28,11 @@ %% worker process out of the pool. Nested jobs are always executed %% immediately in current worker process. %% +%% 'single' mode is offered to work around a bug in Mnesia: after +%% network partitions reply messages for prior failed requests can be +%% sent to Mnesia clients - a reused worker pool process can crash on +%% receiving one. +%% %% Caller submissions are enqueued internally. When the next worker %% process is available, it communicates it to the pool and is %% assigned a job to execute. If job execution fails with an error, no |