Split instantaneous_repl_total_kbps

[DarrenJiang13]

A supplement to #10062
Split instantaneous_repl_total_kbps to instantaneous_input_repl_kbps and instantaneous_output_repl_kbps.

Work:

This PR:

• delete 1 info field:
  • instantaneous_repl_total_kbps
• add 2 info fields:
  • instantaneous_input_repl_kbps / instantaneous_output_repl_kbps

Result:

• master

total_net_input_bytes:26633673
total_net_output_bytes:21716596
total_net_repl_input_bytes:0
total_net_repl_output_bytes:18433052
instantaneous_input_kbps:0.02
instantaneous_output_kbps:0.00
instantaneous_input_repl_kbps:0.00
instantaneous_output_repl_kbps:0.00

• slave

total_net_input_bytes:18433212
total_net_output_bytes:94790
total_net_repl_input_bytes:18433052
total_net_repl_output_bytes:0
instantaneous_input_kbps:0.00
instantaneous_output_kbps:0.05
instantaneous_input_repl_kbps:0.00
instantaneous_output_repl_kbps:0.00

[DarrenJiang13]

@soloestoy @oranagra
could you please check this PR and I will update redis-doc after this being merged.

[madolson]

Strongly against deleting the existing field. We have ops tools that rely on printing the instantaneous metric, I am sure many others do as well. I have no real objection to adding 2 more though.

[soloestoy]

Strongly against deleting the existing field. We have ops tools that rely on printing the instantaneous metric, I am sure many others do as well. I have no real objection to adding 2 more though.

@madolson I can't understand your word, the instantaneous_repl_total_kbps have not been released, why you rely on an unreleased metric? it doesn't make sense.

[madolson]

@soloestoy You're right, I copied the wrong variable name.

We expect only one of these numbers to be non-zero for the most part right?

[soloestoy]

We expect only one of these numbers to be non-zero for the most part right?

You mean instantaneous_input_repl_kbps and instantaneous_output_repl_kbps?

In fact we build the replication chain master->replica->replica->... for users, to avoid heavy replication output flow in master, so the mid replicas' input and output replication flow are non-zero for the most part.

[madolson]

I forgot about chain replicas. I suppose if we are going to have repl_total_kpbs, I have no objection to splitting it between the two.

[madolson]

I'm OK with the change.