From 1f248667c59b832f6e3dd68fc5ee1cdb0f217f8f Mon Sep 17 00:00:00 2001 From: Jan Date: Wed, 5 Dec 2018 13:06:10 +0100 Subject: [PATCH] Doc - additional 3.4 related info (#7645) --- .../Manual/ReleaseNotes/UpgradingChanges34.md | 26 +++++++++++++++++++ 1 file changed, 26 insertions(+) diff --git a/Documentation/Books/Manual/ReleaseNotes/UpgradingChanges34.md b/Documentation/Books/Manual/ReleaseNotes/UpgradingChanges34.md index ef2133d9f3..25bda0b0c2 100644 --- a/Documentation/Books/Manual/ReleaseNotes/UpgradingChanges34.md +++ b/Documentation/Books/Manual/ReleaseNotes/UpgradingChanges34.md @@ -611,6 +611,11 @@ For arangod, the following startup options have changed: The default value for this option is *false*. + As mentioned above, ArangoDB 3.4 changes the default value of the configuration option + `--rocksdb.total-write-buffer-size` to about 40% of available physical RAM, and 512MiB + for setups with less than 4GiB of RAM. In ArangoDB 3.3 this option had a default value + of `0`, which meant that the memory usage for write buffers was not limited. + Permissions ----------- @@ -650,6 +655,25 @@ should change the protocol from SSLv2 to TLSv12 if possible, by adjusting the value of the `--ssl.protocol` startup option. +Replication +----------- + +By default, database-specific and global replication appliers use a slightly +different configuration in 3.4 than in 3.3. In 3.4 the default value for the +configuration option `requireFromPresent` is now `true`, meaning the follower +will abort the replication when it detects gaps in the leader's stream of +events. Such gaps can happen if the leader has pruned WAL log files with +events that have not been fetched by a follower yet, which may happen for +example if the network connectivity between follower and leader is bad. + +Previous versions of ArangoDB 3.3 used a default value of `false` for +`requireFromPresent`, meaning that any such gaps in the replication data +exchange will not cause the replication to stop. 3.4 now stops replication by +default and writes according errors to the log. Replication can automatically +be restarted in this case by setting the `autoResync` replication configuration +option to `true`. + + Mixed-engine clusters --------------------- @@ -714,6 +738,8 @@ This change will be visible only on systems which allow assigning names to threads. + + Deprecated features ===================