which parameter can reduce the count of snapshot?

[huangxuelun]

my garnet version is v2.1.1, and run in .net 10.
below is my settings.
"Port" : 6379,
"Address" : "0.0.0.0",
"EnableStorageTier": true,
"MemorySize" : "2g",
"LogDir": "./logs",
"CheckpointDir": "./checkpoints",
"EnableAOF": true,
"EnableFastCommit": true,
"Recover": true,
"CompactionType": "Scan",
"CompactionFrequencySecs" : 60,
"CompactionMaxSegments" : 2,
"CompactionForceDelete": true,
"AofSizeLimit": "1g",
"EnableIncrementalSnapshots": true

[John Karam (johnkaram02)]

From the screenshot, that's one checkpoint split into 1 GB files (snapshot.dat.0 to .12, about 12.6 GB total). So the thing to shrink is how much data the snapshot holds, not the number of files. I think two of your settings are being silently ignored on 2.1.1, and that's why it's so large.

1. MemorySize isn't a 2.x option. In Garnet 1.x the key was MemorySize. In 2.x it's LogMemorySize, with a default of 16g (libs/host/defaults.conf and Options.cs at v2.1.1: [Option('m', "memory")] public string LogMemorySize). The JSON config is loaded with System.Text.Json, which skips unknown properties, so "MemorySize": "2g" is dropped without any warning. I ran garnet-server 2.1.1 with "MemorySize" in the config file and it started normally, with no error or warning. Your main log is therefore allowed to keep up to 16 GB in memory, and a snapshot checkpoint writes the in-memory part of the log. That fits your ~12.6 GB.

"LogMemorySize": "2g",

2. EnableIncrementalSnapshots doesn't exist in 2.x either. It's in the 1.x defaults.conf (v1.1.10) but not in 2.1.1's, so it's ignored as well. In 2.x, each checkpoint is a full snapshot of the in-memory region.

What controls the rest:

• SegmentSize (default 1g) is the size of each snapshot.dat.N file. Raising it gives you fewer, bigger files for the same total; it doesn't reduce disk usage.
• How often snapshots are taken: you have EnableAOF with "AofSizeLimit": "1g", and 2.x has a background task (AofSizeLimitEnforceFrequencySecs, every 5 s) that takes a checkpoint whenever the AOF reaches that limit. A bigger AofSizeLimit means fewer checkpoints (but a longer AOF to replay on recovery).
• Compaction: 2.x's own defaults.conf describes Scan as "NOT RECOMMENDED" because it builds a temporary index proportional to the keyspace, and recommends Lookup for production. That doesn't change the snapshot size, but it matters with CompactionFrequencySecs: 60.

A good way to check what the server is actually using is to start from the 2.1.1 defaults.conf and edit that, rather than carrying over a 1.x config. Every key in it is one the 2.x loader recognizes.