Skip to content

[BUG]replicationFeedSlaves still high CPU utilization in redis7.0 when it Replication backlog and replicas use one global shared replication buffer #10428

@xiaozhitaba

Description

@xiaozhitaba

Describe the bug
here is valgrind CPU between redis6.x and redis7.x when execute replicationFeedSlaves.

in redis6.x copy argv&argc to echo replica
redis-server
replicationFeedSlaves
67.42%
(1.44%)
1040385×

in redis7.x use one global shared replication buffer,
redis-server
replicationFeedSlaves
58.22%
(0.67%)
4099731×

redis-server
feedReplicationBuffer
56.16%
(9.35%)
41089161×

To reproduce

replicationFeedSlaves funcation still high flow rate cpu 。 when master have 10 Slaves ,

Expected behavior

replicationFeedSlaves CPU utilization < 10%

Additional information

redis-server.valgrind2.out.svg
redis-server.valgrind.buf.out.svg

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

Status

Done

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions