mirror of https://gitee.com/bigwinds/arangodb
33 lines
1.4 KiB
Markdown
33 lines
1.4 KiB
Markdown
Upgrading to ArangoDB 3.4
|
|
=========================
|
|
|
|
Please be sure that you have checked the list of
|
|
[incompatible changes in 3.4](../../ReleaseNotes/UpgradingChanges34.md)
|
|
before upgrading.
|
|
|
|
Upon upgrading from 3.3 to 3.4, the following storage engine-specific data conversion tasks
|
|
will be executed:
|
|
|
|
- **MMFiles storage engine:**
|
|
|
|
All collection datafiles will be rewritten into a
|
|
new data format. This data format is required in order to support using the collections
|
|
in [ArangoSearch Views](../../Views/ArangoSearch/README.md) introduced in ArangoDB 3.4.
|
|
|
|
The conversion will read each datafile sequentially and write out a new datafile in the
|
|
new format sequentially. This means the disk will be involved, but both reading and
|
|
writing are done in a sequential fashion. Preliminary tests have shown that it will need
|
|
at most 2-3 times as long as it takes to copy the database directory.
|
|
|
|
- **RocksDB storage engine:**
|
|
|
|
All existing geo indexes will be rewritten into a new
|
|
data format. This data format is required for using the indexes with the improved
|
|
[geo index feature](../../Indexing/Geo.md) in ArangoDB 3.4.
|
|
|
|
Preliminary tests have shown that the conversion can process about 500K to 1M geo index
|
|
entries per second on commodity hardware.
|
|
|
|
If you upgrade without any existing data (new blank data folder), then none of these tasks
|
|
needs to be run because the datafiles will be created using the new format already.
|