This document gives some information about compaction.
Introduction to compaction¶
In Nebula Graph, compaction is the most important background process. Compaction has an important effect on performance.
Compaction reads the data that is written on the hard disk, then re-organizes the data structure and the indexes to make the data easier to read. The read performance can increase by times after compaction. Thus, to get high read performance, trigger compaction manually when writing a large amount of data into Nebula Graph. Note that compaction leads to long time hard disk IO, we suggest that you do compaction during off-peak hours (for example, early morning).
Nebula Graph has two types of compaction: automatic compaction and full compaction.
Automatic compaction is done when the system reads data, writes data, or the system restarts. Automatic compaction is enabled by default.
Full compaction enables large scale background operations for a graph space such as merging files, deleting the data expired by TTL. Use these statements to enable full compaction:
nebula> USE <your_graph_space>; nebula> SUBMIT JOB COMPACT;
The preceding statement returns a job_id. To show the compaction progress, use this statement:
nebula> SHOW JOB <job_id>;
Do the full compaction during off-peak hours because full compaction has a lot of IO operations.
These are some operation suggestions to keep Nebula Graph performing well.
- After data import is done, run
SUBMIT JOB COMPACT.
SUBMIT JOB COMPACTperiodically during off-peak hours, for example, early morning.
- To control the read and write traffic limitation for compactions, set the following parameter in the
# Limit the read/write rate to 20MB/s. --rate_limit=20 (in MB/s)
Q: Can I do full compactions for multiple graph spaces at the same time?
A: Yes, you can. But the IO is much larger at this time.
Q: How much time does it take for full compactions?
rate_limit is set to
20, you can estimate the full compaction time by dividing the hard disk usage by the
rate_limit. If you do not set the
rate_limit value, the empirical value is around 50 MB/s.
Q: Can I modify
A: No, you cannot.
Q: Can I stop a full compaction after it starts?
A: No you cannot. When you start a full compaction, you have to wait till it is done. This is the limitation of RocksDB.