mirror of https://gitee.com/bigwinds/arangodb
64 lines
3.1 KiB
Markdown
64 lines
3.1 KiB
Markdown
Incompatible changes in ArangoDB 3.3
|
|
====================================
|
|
|
|
It is recommended to check the following list of incompatible changes **before**
|
|
upgrading to ArangoDB 3.3, and adjust any client programs if necessary.
|
|
|
|
The following incompatible changes have been made in ArangoDB 3.3:
|
|
|
|
* AQL: during a traversal if a vertex is not found, arangod will not log an error and
|
|
continue with a NULL value, but will instead register a warning at the query and
|
|
continue with a NULL value.
|
|
|
|
If a non-existing vertex is referenced from a traversal, it is not desirable to log
|
|
errors as ArangoDB can store edges pointing to non-existing vertices (which is perfectly
|
|
valid if the low-level insert APIs are used). As linking to non-existing vertices
|
|
may indicate an issue in/with the data model or the client application, the warning is
|
|
registered in the query so client applications have access to it.
|
|
|
|
* ArangoDB usernames must not start with the string `:role:`.
|
|
|
|
* The startup configuration parameter `--cluster.my-id` does not have any effect in 3.3.
|
|
For compatibility reasons, ArangoDB 3.3 will not fail on startup if the option is
|
|
still used in the configuration, but it will silently ignore this option.
|
|
|
|
* The startup configuration parameter `--cluster.my-local-info` is deprecated now.
|
|
Using it will make arangod log a warning on startup.
|
|
|
|
* Server startup: the recommended value for the Linux kernel setting in
|
|
`/proc/sys/vm/max_map_count` was increased to a value eight times as high as in
|
|
3.2. arangod compares at startup if the effective value of this setting is
|
|
presumably too low, and it will issue a warning in this case, recommending to
|
|
increase the value.
|
|
|
|
This is now more likely to happen than in previous versions, as the recommended
|
|
value is now eight times higher than in 3.2. The startup warnings will look like
|
|
this (with actual numbers varying):
|
|
|
|
WARNING {memory} maximum number of memory mappings per process is 65530, which seems too low. it is recommended to set it to at least 512000
|
|
|
|
Please refer to [the Linux kernel documentation](https://www.kernel.org/doc/Documentation/sysctl/vm.txt)
|
|
for more information on this setting. This change only affects the Linux version of ArangoDB.
|
|
|
|
|
|
Client tools
|
|
------------
|
|
|
|
* The option `--recycle-ids` has been removed from the arangorestore command.
|
|
Using this option could have led to problems on the restore, with potential
|
|
id conflicts between the originating server (the source dump server) and the
|
|
target server (the restore server).
|
|
|
|
* The option `--compat` has been removed from the arangodump command
|
|
and the `/_api/replication/dump` REST API endpoint.
|
|
In order to create a dump from an ArangoDB 2.8 instance, please use an older
|
|
version of the client tools. Older ArangoDB versions are no longer be supported by
|
|
the arangodump and arangorestore binaries shipped with 3.3.
|
|
|
|
Miscellaneous
|
|
-------------
|
|
|
|
The minimum supported compiler for compiling ArangoDB from source is now g++ 5.4
|
|
(bumped up from g++ 4.9). This change only affects users that compile ArangoDB on
|
|
their own.
|