From cbc0d0cf824faab1ba7cda1793caadc937d0128b Mon Sep 17 00:00:00 2001 From: Nico Cartron Date: Thu, 9 May 2019 16:19:24 +0200 Subject: [PATCH] made the 'nometasync' sync mode explanation more clear --- docs/backends/lmdb.rst | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/backends/lmdb.rst b/docs/backends/lmdb.rst index 4bc8e151f8..d63d073575 100644 --- a/docs/backends/lmdb.rst +++ b/docs/backends/lmdb.rst @@ -47,9 +47,9 @@ Default is 2 on 32 bits systems, and 64 on 64 bits systems. Synchronisation mode: nosync, nometasync, mapasync -* ``nosync``: Don't flush systems buffers to disk when committing a transation. +* ``nosync``: don't flush systems buffers to disk when committing a transation. This means a system crash can corrupt the database or lose the last transactions if buffers are not yet flushed to disk. -* ``nometasync``: Flush the data on a commit. Slightly faster than doing a full sync, but can potentially lose the last committed transaction if the operating system crashes. +* ``nometasync``: flush system buffers to disk only once per transaction, omit the metadata flush. This maintains database integrity, but can potentially lose the last committed transaction if the operating system crashes. * ``mapasync``: use asynchronous flushes to disk. As with nosync, a system crash can then corrupt the database or lose the last transactions. -- 2.47.2